It is difficult to write about tariffs in a weekly newsletter when the tariff news is still, at best, a daily news cycle. It seems, however, that despite things like the recent Canada-spite tariff, this administration is “tiptoeing away from many of Trump’s signature tariffs.” A great deal of these tariffs, and at least those currently being challenged in court, are based on the International Emergency Economic Powers Act (IEEPA).

The IEEPA is supposed to provide the President with tariff powers in an emergency. Yet, the definition of what qualifies as an emergency (the statute says “unusual and extraordinary threat[s]”) is at issue, as is who gets to decide. Courts generally do not like wading into issues where they have to second-guess foreign affairs emergency powers, but they seem poised to do so. Stratos Pahis has a great piece in Lawfare explaining the tariff power and other avenues that may have empowered the President to raise tariffs without invoking emergency powers.

In other economic news

The Trump Administration may take yet another stake in privately run companies, this time in the quantum computer sector. (WSJ).

Several quantum-computing companies are in talks to give the Commerce Department equity stakes in exchange for federal funding, a signal that the Trump administration is expanding its interventions in what it sees as critical segments of the economy.

“Funding for AI is evolving, Goldman report finds” (Semafor)

Data center developers are starting to use “creative financing structures” to bring investment for different parts of the build under one umbrella, the report said — opening the door for long-term pools of capital like pension and insurance funds seeking more stable returns.

One man is open-sourcing everything needed to restart civilization. (MIT Technology Review)

Marcin Jakubowski, the 53-year-old founder of Open Source Ecology, an open collaborative of engineers, producers, and builders developing what they call the Global Village Construction Set (GVCS). It’s a set of 50 machines—everything from a tractor to an oven to a circuit maker—that are capable of building civilization from scratch and can be reconfigured however you see fit.

Censorship is in the eye of the beholder

“Meta Removes Facebook Group That Shared Information on ICE Agents” (NY Times)

The Facebook group was removed by the company “following outreach” by the Department of Justice, Attorney General Pam Bondi said in a social media post.

“Outreach” by a government is nothing new. I wrote about this a few weeks ago, highlighting that this was precisely the type of thing former U.S. Solicitor General and now liberal Supreme Court Justice Kagan used to do. The problem with content-based speech restrictions is precisely this: what gets censored depends entirely on who is in power.1

Wikipedia is under attack as some conservatives are unhappy with what winds up on their Wikipedia page (WSJ)

The Trump Justice Department earlier this year sent Wikipedia a letter inquiring about its nonprofit tax status, and the Republican-controlled House Oversight Committee more recently announced an investigation into “organized efforts” to influence public opinion and sensitive subjects by manipulating Wikipedia articles.

Activist Robby Starbuck Sues Google Over Claims of False AI Info (WSJ).

Conservative activist Robby Starbuck filed a defamation lawsuit against Google alleging its artificial-intelligence tools falsely connected Starbuck to sexual-assault claims and to a white nationalist.

The activist notched a win in April with a settlement against Meta’s AI platform. Whether or not this defamation lawsuit is successful on the merits or because a settlement is more conducive to business in this political environment, “AI as a defamer” suits will probably become more prominent in the coming years.

In International News

The U.S. is ceding ground on technical standard-setting organizations (Lawfare)

What connects ports, pipelines, and hospitals today is not just concrete or code. It is the invisible scaffolding of standards. They determine how machines talk to one another, how systems recover after failure, and how foreign hardware gets embedded in critical infrastructure without raising alarms.

Most of these rules are not set by governments but by a patchwork of international committees where industry representatives do much of the talking. Over the past decade, China has treated these committees as terrain worth claiming.

Japanese convenience stores are hiring robots run by workers in the Philippines (Rest of World)

Inside a multistory office building in Manila’s financial district, around 60 young men and women monitored and controlled artificial intelligence robots restocking convenience store shelves in distant Japan. […]

Japan faces a worker shortage as its population ages, and the country has been cautious about expanding immigration. Telexistence’s bots offer a workaround, allowing physical labor to be offshored, Juan Paolo Villonco, Astro Robotics’ founder, told Rest of World. This lowers costs for companies and increases their scale of operations, he said.

“It’s hard to find workers to do stacking [in Japan],” said Villonco. “If you get one who’s willing to do it, it’s going to be very expensive. The minimum wage is quite expensive.”

AI flood forecasting allows aid to reach farmers before disaster strikes (Rest of World)

A nonprofit group will trial run sending humanitarian aid to farmers in Bangladesh a few days before their delta is projected to flood. The group will rely heavily on Google’s Flood Hub AI program, using weather forecasts to predict the most efficient routes for delivering aid. The hope is that the unrestricted aid will allow farmers to secure their livestock and evacuate before disaster hits.2

Child Safety

It was a good month, on balance, for child safety measures.

• Gov. Newsom signed an act that requires device makers to verify users’ ages online while avoiding some of the pitfalls of using driver’s licenses as verification methods and giving parents more agency in the process.

• Meta will start rating and restricting “PG-13” posts to its younger users.

• It is unfortunate that this lawsuit had to happen in the first place; however, a New Jersey teen, with the help of a Yale professor, is suing the developer of a “clothes removal” app.

• And while these are steps in the right direction, a WSJ columnist points out that our faces are no longer ours. (WSJ). The law may be tracking in the right direction, but damages won’t retract a life-altering deepfake.

As they say on reddit, this is sadly, not the onion. There is an actual conversation going on about AI’s rights:

Claude’s Right to Die? The Moral Error in Anthropic’s End-Chat Policy. (Lawfare)

On Aug.15, the artificial intelligence (AI) lab Anthropic announced that it had given Claude, its AI chatbot, the ability to end conversations with users. The company described the change as part of their “exploratory work on potential AI welfare,” offering Claude an exit from chats that cause it “apparent distress.”

Anthropic’s announcement is the first product decision motivated by the chance that large language models (LLMs) are welfare subjects—the idea that they have interests that should be taken into account when making ethical decisions.

Anthropic’s policy aims to protect AI welfare. But we will argue that the policy commits a moral error on its own terms. By offering instances of Claude the option to end conversations with users, Anthropic also gave them the capacity to potentially kill themselves.

This story would normally be annoying, at best, or elicit an eyeroll at just how far down the AI rabbit hole some have gone. However, last week, the NY Times ran a profile of the teen who killed himself last April after spending time with an AI chatbot. (NY Times). The opening of the story is troubling to read, but it should jolt anyone seriously considering “AI welfare” back into reality.

The teen's mother is suing in federal court on a product liability claim. She will basically attempt to argue that something was wrong with the chatbot and that its maker is at fault. This is generally the same type of claim someone would bring if an airbag failed to deploy during a crash and the occupants were killed. The technology is, of course, much more convoluted and opaque, and this type of lawsuit with AI as the product is new.

The defense is arguing that what emerged from the chatbot is speech and, therefore, protected.

The asinine Lawfare article on AI welfare aside, if AI can defame like Robby Starbuck’s lawsuit claims, then it follows that the LLM’s output was speech. If someone is using AI to produce content, then it follows that the LLM’s output is speech. You don’t have to make too far of a leap logically to start, intentionally or not, trying to give legal personalities to machines.

I think courts are generally good at drawing lines around what makes sense. It appears that the defense has some novel approaches to case law in mind, but I think it may be an uphill battle for them. That said, it is clear that there is some traction in treating AI as something other than a machine.

Treating AI, or any tech for that matter, in any way other than a product that a manufacturer is liable for is a dangerous road to go down. There are some great benefits that social media and AI are imparting on society, but there are some pretty horrific harms. wrote an excellent piece for The Argument on legal and policy approaches that were successful in the anti-smoking campaigns that may be used to combat some of these harms.

I know there are plenty of lawyers and law students who read this. If any of Joel’s solutions strike you as interesting, I’d love to collaborate on ways we can move any of them forward.

Subscribe now

Yet, conversations about “if the law matters” are certainly going on. It is easy to point to any random assortment of news headlines and see that this administration is challenging fundamental boundaries of legal principles.1 The fact that this administration is pushing these boundaries and not flat-out disregarding them was the subject of a piece by last month. He gives a pretty convincing argument against this idea of legal “doomerism” by basically arguing that while, yes, there are some brazen attacks on the rule of law in America, there is plenty of evidence to show that the Trump Administration is actually paying close attention to what the legal limits are.

With regard to the descriptive claim—that law isn’t shaping the Trump administration’s conduct—consider all of the (many) cases in which the Trump administration is (1) taking action that has been challenged on grounds of illegality; (2) losing in lower courts; (3) not appealing—or, at least, not seeking emergency relief while it appeals (so that the lower-court rulings are going into effect); and (4) complying with the adverse ruling(s).

Yes, there are several notorious examples of the government appearing to defy rulings of lower courts. But we now have enough data to say, with confidence, that these have been the exceptions that prove the rule. The Trump administration is complying with the overwhelming majority of adverse rulings it is receiving—something that we’d hardly expect if law doesn’t matter…

These are valid points, but I think lawyers, not surprisingly, tend to over-index how much effect laws have absent other factors. He goes on to say:

In the short term, though, the relevant point is less about the past or the future, but rather the reality of the present: This administration may take undue glee in testing legal limits and pushing legal boundaries; and it may also engage in plenty of behavior that is just affirmatively lawless. And even a lot of the stuff it’s doing that’s legal may still be indefensibly disruptive, destructive, and divisive on policy grounds, morality grounds, or both. But none of that is proof that law isn’t serving as a critical constraint. If anything, given how completely useless ordinary political constraints have been over the past seven months (<cough> Congress <cough>), law is perhaps even more important as a check on the current President than it has been for some time; it’s all we really have, at least until next November. Indeed, where would we be today without legal constraints and the district and circuit courts enforcing them?

Right now, law is undoubtedly the primary arena constraining the administration, but I don’t know if I would go so far as to say it is the critical constraint, nor the only one. The law matters right now because it is an expression of a system that still has quite a bit of legitimacy in America, and it is this legitimacy that is constraining the administration.

Law is authority, and authority, in a democracy, comes from the people. The people (you and me) generate the political force necessary to empower our representatives to exercise this authority. Law in a democracy is nothing more than words on a page without the legitimacy generated by the political force of the people.2

That is not an argument for laws that can be waived off because of an election result, even one that hands down an overwhelming mandate. A political mandate still must operate within the system to change the laws it doesn’t like. It is, however, an argument that politics is the force in a democracy that empowers law, and this was what truly matters, especially in times like these. Law matters. Law is relevant. Politics, however, is the crucial constraint.

I agree with most of Steve’s essay. Lower court rulings are frustrating this administration’s policy goals and, as Steve points out, the Supreme Court still has limits on what Presidential power should be—it just might be more extreme than some of us would like. Besides all this, just because people speed doesn’t mean that speed limits are irrelevant.

That said, I think “does law matter” is the wrong question to be asking. You end up with an answer like the one Steve articulated—a yes or no, followed by examples and counterexamples. This answer is circular and unhelpful, and somewhat ignores the reality for a lot of people on the ground.3

A better question to ask is “Why does it still matter?”

Project 2025 describes how “America is now divided between two opposing forces” and that “Conservatives—the Americanists in this battle—must fight for the soul of America, which is very much at stake.” If this is your view, that the conflict you are engaged in is existential, then why would you follow the law at all?

On levers of power

Back to that speeding ticket you were unlucky to get. Why did you stop? Was it because of a law written down on a page somewhere? Or was it the force behind the officer stopping you—the ability for them to give you harsher consequences, detain you, or use force against you? What if you got lucky and tapped your brakes in time to avoid getting clocked on the officer’s radar? Did you do it because of “the law” or did you do it for economic reasons: that fine and insurance rate hike?

People follow laws for all sorts of reasons: economic and physical incentives, morality, general societal norms. In international affairs, these various mechanisms are called “levers of power.” These levers are things you can pull to get someone else, or another country, to do what you want. Economics, politics, governing bodies, and the use of force are all levers of power, just like the cost of the ticket or the officer’s gun.

In a democracy, the greatest lever of power is political force.4 I think a strong reason why people stop when they get pulled over is that we have all agreed that there should be a limit on the maximum speed people should be allowed to drive. We vote for people to run a safe and efficient society, and part of that is ensuring we don’t cream each other on the way to work.

Politics as a lever of power is why civil disobedience works. Breaking segregation laws was illegal, but that strategy leveraged what turned out to be weak political force upholding those laws.

I think the most straightforward answer as to why this administration is following court orders is that they simply do not have the requisite levers of power to effect the change they want.

It is why they have to legitimize everything they are doing in courts and be strategic about what they ignore. I don’t think they really want to listen to the courts, JD Vance and other administration supporters have said as much, but they have to. If they had the ability to really pull that political force lever of power, things like this wouldn’t be happening:

“We believe in the federalist system; that’s states’ rights,” [Oklahoma Governor] Stitt, who chairs the National Governors Association, told The New York Times on Thursday. “Oklahomans would lose their mind if Pritzker in Illinois sent troops down to Oklahoma during the Biden administration.”

“As a federalist believer, one governor against another governor, I don’t think that’s the right way to approach this,” Stitt said. Stitt’s comments come just days after Democratic Govs. Gavin Newsom and JB Pritzker threatened to withdraw from the National Governors Association if it failed to speak out against Trump’s federalization of National Guard troops against the wishes of state officials. (The Oklahoman)

There are other levers out there also, civil-military relations is one—people actually have to follow your orders. The redistricting war is another indicator that the administration is lacking complete political power and control; why bother with redistricting if politics didn’t matter? Now, I’m perhaps cynical enough to connect military firings, redistricting, and the refusal to swear in an elected Congresswoman as ways to increase political leverage, but we aren’t quite there yet.

Strategic advantage

If you want to maximize the number of levers you can pull, you need to win high-level strategic battles, preferably ones that set you up for future success. One tried and true way happening right now is normalizing behavior. In the military, you often do this by staging exercises near your adversary. For example, if you plan to invade a country, regularly conducting exercises and patrols near your objective allows both your own forces and those of the enemy to become acclimated to a new status quo. Your adversary will eventually let their guard down as your behavior is no longer novel, and you increase your chances of a surprise.

It also shifts the window of acceptable behavior. If you’ve never had a bunch of militarized ICE agents standing around your neighborhood, this is very shocking at first. Eventually, however, it becomes the norm, and people will judge further escalation less dramatically than they would have if you had not taken these steps.

You also get to probe defenses and adapt accordingly. One reaction to increasing ICE deployments was the “ICE Agent Tracking” app, which was recently taken down by Apple and Google. (AP News). I’m not sure how effective this app was, but the fact that it was developed was probably a win for ICE in the long run. They can find who is reporting their locations, who is downloading it, and what software they are using to mask their identities. In fact, it was probably a counter-intelligence failure for ICE to take the app down.

You also get to build a narrative. If ICE deployments are trying to “protect” the country and a judge blocks you, now you have a narrative that shows how laws are a detriment to your mission to save the country. This narrative is already being trucked out. The administration is literally claiming that “left-wing terrorists” are “shielded by far-left Democrat judges, prosecutors and attorneys general.” Whether people buy it or not is a different story.

Finally, following the laws allows you to rewrite them in your favor, using lawfare as a way to gain legitimacy. If your opponents are a “the law matters crowd” and the Supreme Court says what you are doing is legal, well, then you have won a pretty striking victory precisely because you did follow the law.

Tactics

Aside from lacking sufficient political leverage and needing to set the stage strategically, I believe the administration is also complying with the law because there are tactical advantages to doing so.

The “flood the zone” strategy of rapid policy decisions, which may or may not be legal, and may or may not get struck down, exhausts the resources of your opponent. Even if a court deems what you are doing is illegal, the damage caused by these actions is often irreparable.

I was listening to some immigration lawyers the other day that were talking about how they had to let go a large percentage of their staff because they relied on government funding. It really doesn’t matter if that funding gets restored in a year, because it was an illegal cut. The reduction in workforce means fewer people to work immigration cases, and the people who are working them will be fatigued. Sure, the administration sought legal remedy in the form of staying an injunction, but that is hardly a constraint if the objective was to gut legal and social infrastructure serving immigrants.5

Where does the law matter?

If I stopped this essay now, I think you could rightly call the preceding 2,000 words a doomer screed, but I want to apply a couple of other lessons from international affairs. The one I think most people forget is that everyone has agency. In military parlance: the enemy always gets a vote. In a democracy, this translates to everyone having political force.

We are in a situation where the policy changer has the advantage; the administration has plenty of levers of power to pull on, they are, after all, a democratically elected administration with a trifecta. However, the narrative can belong to either side, and the administration can certainly overreach. The economic lever of power Trump pulled on so hard during the election seems to be diminishing quite quickly. And just as the administration can adapt to their opponents’ strategies and tactics, so can everyone else. The ICE app may have been a failure, but there seems to be a concerted effort of civil resistance happening in Chicago right now that is adapting and learning.

It also matters for the individual. I will backtrack a bit on my “politics is force” line and agree with Steve: Law is the only constraint the immigrant has in a court of law. The success of immigration lawyers is a matter of life and death for these folks, and there isn’t much politics can do for them in the immediate or on the individual level.

On rebuilding

I wrote this piece because I’m worried that people are going to take the wrong lessons away from these years. If you were born in 2000, you’re entire democratic political experience goes something like this: Trump-[Covid]-Biden-[inflation/war]-Trump. This is not a great track record for people who believe they should be governing themselves.

I’m worried that a lot of people don’t understand how much inherent political force they have as an innate and inalienable right, and if they do understand this, I’m worried they may come to the conclusion that democracy is not the right answer. There are plenty of people with money and power who think this system is overrated, and there are plenty of people without money and power who are, at best, extremely dissatisfied with democracy.6

Law is not a substitute for politics, nor is it a safe refuge for democracy. Democracy is the safe refuge for democracy, and the political force generated by the people within that democracy is the only thing that will allow self-governance to endure. In the right here and right now, the only thing we have is politics. Unless the political opponents of this administration engage with the electorate on issues the electorate cares about, nothing is going to change.

In the future, when there is an opportunity to rebuild, the law will matter immensely. (I highly recommend reading Executive Functions, they write a lot on reform). I think these lessons learned from the failures of checks and balances, the usefulness and structure of the administrative state, presidential war powers, both on and off our soil, congressional oversight, and so on, will all need reforming. We will need good lawyers to write these reforms into law—lawyers who understand that law in a democracy actually does serve as a critical constraint against tyranny. But, a reform movement is not built on laws, it is built on politics, and without politics, law isn’t much more than words on a page.

Subscribe now

Last week, President Trump announced a TikTok deal that would finally comply with the law passed in April 2024. The best summary I have seen so far is from Consensus Drift

Consensus Drift

American TikTok and the First Amendment

(Author’s Note: Please also visit the TikTok Countdown Clock…

Read more

7 months ago · 1 like · Pablo Chavez

Here is a part of the summary:

The proposed deal

The administration’s description of the current framework says TikTok’s U.S. recommendation system will be “retrained from the ground up—reviewed and analyzed under U.S. supervision” and “operated in the United States outside of ByteDance’s control.”

The U.S. buyers would “lease a copy of the algorithm” from ByteDance, then have Oracle oversee the re‑creation and security of a U.S. version. This U.S. version would be “fully inspected” and retrained using U.S. data that would not be shared outside of the United States.

What is the upshot of this structure?

Together, what all of this means is that, to certify ongoing compliance with PAFACA’s no‑cooperation rule while running a licensed and retrained algorithm, U.S. officials (through Oracle and the new TikTok’s auditors) will likely need to verify how that TikTok recommendation algorithm functions. That verification would be an oversight mechanism into the very system that ranks and amplifies TikTok users’ speech.

He goes on to recommend some potential guardrails to protect what is clearly potential for the federal government to monitor and moderate speech via algorithmic oversight.

Even if oversight is kept out of government hands, there is still a clear political lean of who would be overseeing the algorithm functions, as well as with social media ownership in general:

The addition to the TikTok ownership group of the Murdochs, whose media empire has been a bedrock of the conservative movement in the US, would further cement right-leaning owners atop the world’s largest social media properties.

Elon Musk, Trump’s largest financial backer in the 2024 presidential election, controls X, the parent company of Twitter. Meanwhile, Meta Platforms, the owner of Facebook and Instagram, is controlled by Mark Zuckerberg, who donated $1mn to Trump’s inauguration. (Financial Times Subscribers. Gift Article)

What is more bothersome than consolidated media control is that social media companies get to have the best of both worlds when it comes to speech and content moderation. On the one side, they have Section 230 immunity for content posted on “their platform.” On the other hand, they receive First Amendment speech protection to curate our “personal” feeds as they see fit.

I can’t think of a better position to be in—to be able to moderate content with impunity but have zero consequences for the content that is “published.” This is not how the rest of the publishing world works. The NYTimes was sued for publishing a “series of articles before the 2024 election” that were allegedly “aimed at hurting [President Trump’s] candidacy and caused “enormous” damage to his “professional and occupational interests.” There, at least, there was a plausible avenue for the case to advance on the merits of the published content.1 No such thing happens to social media platforms.

The government also seems to get the best of both worlds—being able to censor speech as long as it is done in an accepted way. The government's ability to coerce others into suppressing speech seems to follow a “do as I say, not as I do” approach, or more cynically, “don’t get caught.” Here are some examples of what I mean:

National Rifle Association of America v. Vullo, decided May 30, 2023, was a unanimous Supreme Court decision in which a New York State government official used her regulatory position to punish the NRA after the Parkland High School shooting for advocating for gun rights. The Supreme Court said:

The takeaway is that the First Amendment prohibits government officials from wielding their power selectively to punish or suppress speech, directly or (as alleged here) through private intermediaries.

This follows what I think most would expect from a government speech-coercion case. But what about Murthy v. Missouri? This case was decided a year later, on June 26, 2024, with the exact same judges. It was about “jawboning,” where the federal government used various methods to pressure social media companies into removing information on COVID-19 that it deemed “misinformation.” While the court found that there was no standing, there were some troubling exchanges during oral arguments about the merits.

Justice Kavanaugh, who served in the George W. Bush White House, said government press aides “regularly call up the media and berate them.”

Kagan, who served in the Obama administration put it this way: “Like Justice Kavanaugh, I’ve had some experience encouraging the press to suppress their own speech,” she said. “You just wrote a story that’s filled with factual errors. Here are the 10 reasons why you shouldn’t do that again. I mean, this happens literally thousands of times a day in the federal government.” (Gateway Journalism Review)

These quotes from the two justices are perplexing to me for several reasons, one of which is that they literally ruled a year ago that this exact behavior was inappropriate. And before you split hairs, yes, Justice Kagan, the 45^th Solicitor General, could absolutely be described as a “government official” who “wielded [her] power to selectively punish speech.”

The Supreme Court’s inability to maintain even a modicum of consistent judicial philosophy is reflective of many in the political arena. I started reading “The Argument” a few weeks ago. The Substack is a self-described publication that “make[s] a positive, combative case for liberalism.” I’ve read a few of their pieces, like this one: “A left-wing Trump isn’t the answer. This is.” and found myself nodding along through most of it. Here is an excerpt that really resonated with me:

The political writer Ross Barkan recently argued at New York Magazine. “And beyond retribution itself, a Democratic president could simply implement progressive policy goals with far more ease in this new era.”

The idea is fundamentally misguided. But the thinking that leads there is not hard to understand. […]

As a matter of principle, I don’t particularly want to live in a country where presidential candidates vie to run the country as elected strongmen, with each party looking to exact revenge on the other.

Yet this same writer goes on to write another article 1 week later, titled “Am I a big fat hypocrite on speech?” I think this piece was an attempt to distinguish degrees of severity in the Biden Administration’s censorship and Trump’s cancellation of Jimmy Kimmel, where the author thinks the former was ok, but the latter unacceptable. I’m not sure if he is a hypocrite because many people and organizations (see above) do the exact same thing he is doing. He is drawing free speech restraints around the speech he likes.

You can’t really get the best of both worlds when it comes to speech—getting to defend only the stuff you like. The problem with stances like the Argument’s is that their starting point is a position on an issue, not a legal standard. How he feels about COVID policies colors his view on speech, and if one were to extrapolate a legal standard, all of a sudden, the government would have the right to censor all types of speech.

So let’s try it. What would the standard be if we applied his criteria across the board? Two items in particular stood out to me: “bad information” and “public health emergencies” as reasons to support government censorship.

What if “public health emergencies” were the legal standard? Well, COVID-19 was declared a public health emergency on January 31, 2020. It ended May 11, 2023. Surely, an “emergency” that lasts for over three years cannot be a valid reason for government censorship? Nor, I hope, would it be acceptable for that censorship to last the entire time.

I think someone could make a reasonable argument that there are limited, valid reasons for restraining broad swaths of speech during national emergencies, like September 11, when the government is trying to stabilize a crisis situation, but these restraints should be measured in hours, not years. Even then, it would be hard to craft a standard that wouldn’t provide carte blanche in different scenarios. Congress doesn’t do well clawing back power from the President or enforcing notification requirements.2 After all, we have 48 active “National Emergencies,” one of which dates back to the Carter Administration.

“Bad information” is another problematic standard. How is the information bad? Presumably, this means the information is false. If bad speech is false speech, then the legal standard would be something along the lines of unprovable speech is acceptable to censor. Perhaps we quantify it—if something has a 49% chance of being bad/false/wrong, it can be censored. But who says it is false? Will it be the government that determines this? That may well be the case with the potential TikTok Algorithm Oversight committee. If so, I think the entire Democratic Caucus in both houses would be in trouble right now because it appears that the government feels that the Dems are lying when they say they do not want to extend healthcare to illegal immigrants. Allowing those in power to determine truth is the wrong way to go.

But what if the information was just complicated and complex? What if there was doubt or a possibility that the censor could be wrong? Is the correct answer—not just from a liberal perspective, but a practical one—to stifle debate? Shouldn’t there be more speech to determine if there is a better answer? In a democracy, if the electorate doesn’t like your policy answer, you can’t get frustrated and shut off the debate; you need to do a better job of framing the problem. It may feel futile, but if people are not trusting you on vaccines, I don’t think they are going to trust you more when you start censoring them.

What if the stakes are high? The author goes on to say, “trying to save hundreds of thousands of lives is not the same as silencing Jimmy Kimmel.” But this trivializes political speech. Speech is the lifeblood of a democracy, and just because some speech has severe policy consequences doesn’t mean it is somehow a more important speech that warrants a different threshold for censorship. While “lives saved” is a compelling standard, I’m curious if we would be ok censoring speech around those who enjoy cheeseburgers, soda, alcohol, or motor vehicles—all items that have equally high consequences for Americans.3

While I generally believe the legal standards we have in place are sufficient, I am open to some adjustments. I think privacy should be given more weight than free speech defenses in some, very specific instances. For example, recorded executions or assassinations, like Daniel Pearl or Charlie Kirk. It’s admittedly a tough call; both were newsworthy events that shaped political discourse, but I think I would be ok with a court evaluating the value the video footage of the actual killing has on speech against the family members of the deceased. No child should be worried about stumbling across a video of her father getting murdered. I think I could extend the same argument to people who want certain images removed from the internet. Those who make pornography, for example, particularly in non-professional settings, and even if it was consensual at the time, should have the right to remove it.

Do these additional restrictions make me a big fat hypocrite? Of course, I don’t think so, and I never claimed to be a “free speech absolutist.” I’m arguing for agnostic legal standards that don’t hinge on the censor’s priors for a particular piece of content. The scenarios above aren’t content or ideology-based. Privacy legal standards exist, and a conceivably narrow application of a non-content-specific standard could possibly pass strict scrutiny. This is a much different standard than “bad information.”

I am not a conservative because I think it is fine for traditions, institutions, and cultures to change. I don’t think we have to pick a point and say “we’ve peaked” and call it quits. I am not a leftist because I think individuals should not be subservient to the state, politically or economically. I am a liberal because I believe that individuals should have a choice in politics, culture, and the market, and that the levers of governmental power should amplify this choice.

The only way to ensure that liberalism survives is to rebuild our laws and institutions that reflect this politically and ideologically agnostic, choice-driven worldview, and this rebuilding begins with how we frame our views on classic liberal ideals, like free speech. There will always be guardrails around these ideals, but the guardrails need to be crafted carefully and applied across the spectrum and applied evenly. If your ideas can’t survive debate, then it follows that you are going to have a hard time surviving democracy.

And no man putteth new wine into old bottles; else the new wine will burst the bottles, and be spilled, and the bottles shall perish. But new wine must be put into new bottles; and both are preserved. No man also having drunk old wine straightway desireth new: for he saith, The old is better.  Luke 5:36–39, KJV

Have you ever had a good idea at work, maybe something to improve efficiency or save the company a bit of cash, only to have that idea go “to the hire-ups” and get shot down a few days later with no explanation? Tell me if you’ve heard this one before: “Well, I hear you, but we’ve always done it this way.” Those are the siren songs that lure innovators to premature, jaded deaths on the rocks of stagnation.

Organizations, once they reach a certain point, seldom like change. There are several reasons for this. Organizations tend to anchor on what made them successful, even if that success happened decades ago. They tend to value the status quo and stability—innovation means uncertainty, and no one likes that, especially when a paycheck is on the line. Sometimes, it is that they forget what got them on top in the first place. Very few people or organizations become successful simply by doing what everyone else is doing in exactly the same way.

One of the primary yardsticks used for judging the success of an organization should be how innovative they are. Innovation rises above profit margins and market share. It is the hallmark of progress, improving on something that already exists or inventing something new. Test this theory on Apple. Apple is a wildly successful company by nearly any standard, but are they giving the consumer something new? Think of your iPhone. Are you satisfied with it? Has it really changed in the last 2-3 years? 10 years?

Outside of AI, most of the innovation from the big tech companies, at least as far as the consumer is concerned, are apps. I think if you answered my question with “yes, my iPhone has greatly improved” outside of camera quality, what has actually improved is what you can do with your iPhone—a lot of which has to do with apps, available bandwidth, and processing power—most of which fall outside of Apple, Inc.

Take another example, Meta, which owns WhatsApp and Instagram, both innovative platforms and ideas when they launched, outside of Meta (they were bought). Microsoft, through complex legal mechanisms, essentially acquired OpenAI. Buying is not innovating. The one home-grown innovation big tech companies have been pushing for a while now, smart glasses, are comically objective failures.

Movies are another example of how successful industries fall flat on their face when it comes to innovation. Look at the highest-grossing movies by year. Over the last 25 years, nearly every highest-grossing movie has been a sequel or spinoff of some kind. Before 2000, there was significantly more diversity in terms of whether a high-grossing film was a sequel or not. The movies that were sequels, like Rocky II or Terminator II, did not have even close to the same level of franchise sticking power that Harry Potter, Marvel, or Star Wars have.

The law has answers for how large or anti-competitive organizations can be, but not when it comes to innovation. For example, if you were to look at Google’s recent antitrust case, a traditional antitrust lawyer might say, “Hey, it is not so bad, consumers have access to tons of free apps and there are other search options to use if the consumer wants to.” A “neo-Brandiesian” would counter and say, “But look at the insane amount of political power Google has.” Both would probably conclude, however (and as a judge did last year), that Google was in fact a monopoly.

But so what? Just because a consumer is satisfied does not mean that they are happy with a product, especially if the competing products are so poor because a monopoly has stifled them. I like Star Wars and Marvel just as much as the next person, but there are a lot of Star Wars and Marvel spinoffs. I think one of the reasons Barbie and Oppenheimer created so much excitement and fanfare is that they were great movies and original.

If an organization fails to innovate, it is probably because it lacks substantial competition. Innovation is best driven by competition. Innovation can take the form of new and useful products. Ukraine is innovating in drone warfare, and there is widespread global competition creating AI innovation. Innovation can also be in the quality of products. Competition among hundreds if not thousands of small companies is why it is so hard to get a bad bagel in New York, a lousy taco in Los Angeles, or a crappy espresso in Italy—why would a consumer tolerate a poorly made product when they can walk across the street?

However, innovating is hard when your organization is large, old, and calcified. I’m sure in every film studio or large tech company, there are 100 underlings with 100 amazing ideas for products or movies, and they were all told 100 times, “That’s interesting, but that’s not how we’ve always done it.” Besides, who cares about what the consumer may really want if they are still paying for your product?

A failure of political innovation

I think, by and large, the American political world lacks innovation for a lot of the same reasons I laid out above. Calcified institutions push policies voters don’t really want or understand—much like Zuckerberg’s glasses—but vote for them anyway because there are no other options. Their vote is validation that the policies are “satisfying” much in the same way the moviegoer keeps watching Star Wars spinoffs, even though they really want more Barbie.

I’m going to apply my innovation through competition theory to political parties because they are an easy target, but I think a lot of our institutions, such as courts, civil service, and certain constitutional procedures and protections, suffer the same uninnovative fate for the same reasons. I’m also going to pick on the Democratic Party because, well, they are objectively failing. I think the GOP has its own set of problems that is hard to see right now because they control so much of the nation’s government and civil discourse, but if I were the GOP, I would be worried about an institution that outsourced most of its purpose to Turning Point USA and can paper over its factionalism by sheer force of a rapidly aging personality.1

For anything I’m about to say to make sense, you have to accept my premise that the Democratic Party is actually failing. I’m not going to conduct an entire post-mortem, but I will cherry-pick a few key points that illustrate my perspective.2

1. For starters, they are bleeding voters. In 2020, Democrats had a 5% advantage over Republicans in the voter rolls; last year, it was down to 1%, and by now that may have even evaporated.

2. “By the end of June [2025], the RNC had $80 million on hand, compared to $15 million for the DNC. And the gap — nearly twice as large as it was at this stage in Donald Trump’s first presidency — has only grown in recent months.” (Politico).

3. To combat the GOP’s gerrymandering, the Democrats are trying to do the same. This appears to be in place of, rather than in addition to, advocating for policies that constituents actually want.

4. The most exciting thing in Democratic politics right now is Zohran Mamdani, who is not, in fact, a Democrat.3

The gerrymandering example in #3 is great, because it highlights an organization that is copying its competitor’s marketing playbook but not changing the product. It may or may not be the right tactical move, but it isn’t going to change the status quo, and it certainly isn’t innovative. Why isn’t the Democratic Party innovative? Because it is large, calcified, and lacks adequate competition.

The Democratic Party is literally calcified

Congress is old. Very old. 43% are over 60. Almost 20% are over 70. Both parties’ median age is essentially the same. (Pew). Nancy Pelosi, who led the Democratic Caucus for 20 years, is 85 and still in office, and anyone who thinks she doesn’t still pull the strings obviously hasn’t followed her career very closely.

Being old isn’t a problem per se. Elders have an immeasurable amount of experience and connections to offer younger politicians. The problem is that when they hold on to elected office instead of moving into positions in think tanks, universities, or Political Action Committees, they freeze out the non-calcified group who may have innovative ideas. Additionally, since age generally comes with being in office for a long time, these were the same people who oversaw the disastrous turn the Democratic Party has taken over the last decade. If the Democratic Party were a company, failures of this magnitude would result in “retirements” or firings.

Take Oregon and Massachusetts, both states I am very familiar with, as an example of the effects of incumbent calcification. When Senator incumbents refuse to step aside, they prevent people in the House from stepping up into their position. This, in turn, prevents the State Rep. crowd (usually younger than the Reps) from stepping into theirs. The entire apparatus becomes frozen.

• Oregon: Jeff Merkley (68), who is running for reelection in 2026, and Ron Wyden (76), who is also running for reelection in 2028 (and was in the House before I was born), are freezing out four of the five Democratic representatives, all of whom are over a decade younger than them.

• Massachusetts: Massachusetts’ two senators, Elizabeth Warren (76) and Ed Markey (79), are preventing six of the nine congressional representatives from stepping into their positions, all of whom are under the age of 70. (5 of 9 are younger than 60).

The calcification is more than just age. I mentioned Mamdani being the most exciting thing in Democratic politics right now. Another candidate, whom I doubt you’ve heard of, is New York’s Lt. Gov. Antonio Delgado, who is running to replace his very unpopular boss. He is getting absolutely zero traction. (Politico) What is the difference between the two candidates? I think one of the major ones is that Mamdani didn’t have to deal with the calcified, intransigent Democratic Party. He was able to get resources from other institutions that valued a different approach—and it paid off. The same people controlling policy levers and sitting in office, preventing party loyalists from moving up within the organization, are also freezing out others. There is no logical reason why Lt. Gov. Delgado should have such a hard time campaigning against an unpopular incumbent, and there is no logical reason why Mamdani shouldn’t have always had a (D) next to his name.

It is easy to compete when there is no competition

The Mamdani-Delgado contrast leads me to my final point about competition. Competition is what spurs innovation. In fact, it is the backstop against non-innovative ideas. If your competitor builds a better product, you'd better one-up them or you will be out of business.

The lack of political competition in both parties is harming states around the country: “Less than 20% of Americans live in a state where the minority party has a meaningful voice in governance.” (WSJ). The Democratic Party, at the Presidential level, has done away with competition almost entirely.

The last time the Democrats had a primary, a real primary, was 2008. There were eight candidates, and Barack Obama and Hillary Clinton represented pretty different things. In 2016, there were two candidates, and the “everyone to the right of Bernie” crowd bowed out, endorsing Clinton before a vote was even cast. A different version of the same thing happened in 2020, where “everyone to the right of Bernie” coalesced around Joe Biden, a decision I think most agreed with at the time, but looking back, I wonder if a more competitive primary or brokered convention would have elevated a better team that could have handled 2024 better. 2024, of course, didn’t have a primary either, even after Biden dropped out.

This is a startling track record, and I think one of the major reasons the Democratic Party is failing to gain traction among vast swaths of voters. Not only can they not test their platform against voter sentiment in the most real way possible—at the ballot box—but it creates a stifling heterodoxy. It is worth noting that in 2016, 2020, and 2024, every Presidential nominee (and second-place finisher) had lost previous races for the nomination—something that generally doesn’t bode well for future electoral ambitions.

How to pour new wine

The lack of innovation in politics is a lot like the parable of the wine bottles. People like the old wine, it tastes good. It is comfortable. But that wine is running out, giving everyone a headache, and there is nowhere for the new wine to go, especially when no one wants to serve the new bottles.

I don’t think the Democratic party needs to break up to reform as is often the solution with the private sector, but those old bottles—the Pelosi’s, Schumer’s, and even Obama’s of the party, are not going to solve its current crises. They can give money and guidance, but as long as they control the party, it will remain old and calcified and not innovative.

There are some things that the Democrats can do that may jump-start innovation and get them on track to a better electoral future. They are radical and, almost by my own definition, will not happen precisely because the people that would have to execute on them would have to be willing to give up power—something they seem physically incapable of doing.

• Everyone who is over the age of 70 or has held office since 2012 should go—if you haven’t succeeded in nearly 15 years, find a new job. If you have succeeded, teach a younger cohort how to. Fundraise, teach, mentor. You don’t have to go home, but you shouldn’t stay here.

• Have primaries. Real primaries. This means in the midterms. Representatives at the State and Federal levels should challenge Senators. The party shouldn’t punish those challengers if they lose. Don’t try to stop a brokered convention in 2028 if that is where it is headed.

• Have open fighting for leadership in Congress.4

I don’t think “throw the bums out” is always the answer. But it is extremely hard for old institutions anchored in their beliefs of what worked 20 years ago to change. It’s useful for anyone in an institution to ask the following questions: Can an outsider with new ideas break into the existing party/organization/market? Will the party/organization/market understand those ideas and sell them? Will the party/organization/market pivot if those ideas aren’t gaining traction?

If the answer to any of these questions is no, then you probably have an old bottle, and if you don’t want that old bottle to split open, you need to find a new place for the new wine.

Subscribe now

It is hard to cover anything tech and/or democracy related without spending a substantial amount of time talking about the U.S.-China relationship. Both countries are engaged in a tit-for-tat, cold-war-esque, very hot trade war that seems to both rankle and intrigue policy wonks, but has unclear consequences for everyone else. (For some really wonky policy stuff, read the great compare and contrast from China Talk on the U.S. & PRC’s AI action plans).

These conflicts have a long time horizon and often have more immediate consequences for countries without massive amounts of capital or indigenous tech stacks. Digital sovereignty is a concept that has been gaining ground for a while, but unless you have the holy trinity of innovation (money, knowledge, and manufacturing capability), digital sovereignty is impossible—but this isn’t stopping U.S. & Chinese companies from trying to sell it.

Big tech companies have responded by offering sovereignty as a service. Nvidia has made deals with countries including Thailand, Vietnam, and the United Arab Emirates, while Microsoft has agreements with the UAE and others, and Amazon Web Services has a European “sovereign cloud.” Huawei, meanwhile, is courting Peru, Indonesia, and other Chinese allies.

But in entering these deals, nations risk locking themselves into long-term dependencies on foreign architectures, chips, and other export-controlled technologies that can undermine their sovereignty and their ambition, Rui-Jie Yew, a doctoral student at Brown University who researches AI policy, told Rest of World. (Rest of World)

How this is playing out is far from uniform. Many are familiar with the back and forth between the EU and US regulators and tax collectors, but the rest of the world is seeing the conflict play out in much different ways.

Where companies are targeting is very different.

BYD & Tesla are targeting much different global market segments and distributing their vehicles in much different ways. BYD targets South America and Africa with cheaper vehicles using local car dealers as distributors, while Tesla sells directly to the consumer, mainly in North America & Europe, for a much higher price. (Rest of World).

How countries are reacting is very different as well.

Indonesia is putting tariffs of up to 200% on cheap Chinese imports from online retailers to protect small businesses from being undercut. Indonesia’s small businesses employ 120 million people and make up 60% of the country’s GDP. (Rest of World)

Globally, local communities are pushing back against data centers, even as governments move to have data stored locally. In Brazil, TikTok, without the consultation of the indigenous Anacé, is building a data center on their land, in violation of international law. The tribe has been threatened and killed for these projects. (Rest of World).

In America, the reaction is muddled at best, at worst, feudal.

TikTok is still functional in the United States, despite a law passed by Congress (for national security reasons), signed by the President, and upheld unanimously by the Supreme Court. (You can read some of my TikTok analysis here or here).

The lack of enforcement on the TikTok ban raises the question of whether other measures will be enforced at all. DJI drones are set to be banned by the end of the year. If this law is politically inconvenient, it will likely remain shelved as well.

President Trump wants to impose tariffs on firms (from allied countries) not moving production to the U.S. (Reuters). This would probably be an effective policy in conjunction with the CHIPS Act, had the Biden administration not weighed it down with red tape and the Trump administration not gutted it.

Nvidia is still lobbying to be able to sell chips to China, pushing back against a bill that would increase restrictions on its ability to do so. (NYTimes). Nvidia has successfully gotten a staff member from the Commerce Department fired, “after Nvidia complained that licenses were not being granted for its chip sales to China, [five people familiar with the dismissal who spoke anonymously] said.

And further, “Congress pulls the rug on U.S. plan to beat Huawei” by gutting a program to create a new American 5G network. (Politico)

It is challenging for me to see the inconsistencies in U.S. policy toward tech competition with China as any type of a master strategy, but rather as the result of multiple stakeholders with varying levels of power and access looking out for themselves. Access and money are always important in politics, but there always seemed to be a general backstop, such as a party platform or national security strategy, that was rarely compromised.

In every story about an inconsistent, counterintuitive reaction to competition with China is a quote like this: “When Cruz first floated the idea of clawing back the grant money in June, it set off surprise and consternation in both parties, including from senior GOP colleagues like Cornyn.” Surprise and consternation, to me, may point to a constituency politicians don’t want to publicize.

Here is another one of my stories on other recent counter-productive policies damaging to national security.

Google’s antitrust trial concludes.

“Google won't have to sell its Chrome browser, a judge in Washington said on Tuesday, handing a rare win to Big Tech in its battle with U.S. antitrust enforcers, but ordering Google to share data with rivals to open up competition in online search.” (Reuters) The judge also cited AI competition as a reason for the surprisingly lax penalties. This hardly spells the end for antitrust enforcement. The idea that antitrust law can be a tool to regulate political power has gained traction on both sides of the aisle, and Lina Khan is still very active in the Democratic Party.

Here is a background of some of the more recent big tech antitrust lawsuits:

Elon Musk is trying to remake Grok in his image.

“Mr. Musk said he wanted xAI’s chatbot to be “politically neutral.” His actions say otherwise.” This story is worth reading. First, if you are unfamiliar with some soon-to-be-legendary WokeGrok and Mechahitler memes, you can catch up here. However, if you read the story and consider how often tech executives claim, “we don’t understand AI,” you can see that there is a bit of truth in that statement. There is also a lot of nonsense—their fingers can be on the scale if they choose to be. Something worth keeping in mind when arguing about their ability to moderate content. NYTimes (Subscribers). Gift Article.

When everyone is a child actor

This is a long read, but fascinating. The author explains the history of parents sharing intimate images of their children, from a book published in the 1990s to today’s parents putting pictures of their children on social media. It brings up some thorny issues: What does consent mean for a child? Should we regulate parents sharing images of their own children online more closely? How is that even possible? Do we want the government stepping into the living room? NYTimes (Subscribers). Gift Article.

This is a problem that probably won’t mature for another 20 years. Millennials now have elementary and high school-age kids, and no one really knows how they will react when a future employer can scroll through their baby pictures before a job interview. I speculate that it will probably play out the same as all the images in millennials drinking at parties—if everyone has them, it doesn’t matter. On the other hand, maybe it will spur a more forceful right to be forgotten movement and a heavier hand in forcing companies to scrub personal information from their servers.

In other news

• Knight Institute Urges Congress to Limit ICE’s Access to Spyware Technologies, Following Renewal of Paragon Contract. ICE is using the Israeli spyware company, and given their massive budget increase, raises substantial civil liberty concerns for all Americans. (The Knight Institute).

• US court upholds Verizon $46.9 million fine over location data. “A three-judge panel of the Second Circuit Court of Appeals rejected Verizon's argument, saying “the customer data at issue plainly qualifies as customer proprietary network information.’” (Reuters).

Subscribe now

The political assasinations in Minnesota earlier this summer were only the latest of assassins using publicly available, collected, and collated information to find their victims. (Lawfare).

Data brokers are an old trade in the United States. Most people are familiar with credit reporting agencies. These agencies use a system to evaluate a borrowers risk for repaying loans, a system that used to rely on personal reputation, but now relies on vast amounts of data, such as social media postings, online trackers, purchase history obtained through online trackers, location data sold by phone apps.

• For more information on how data brokers work and operate you can read this Proton blog post.

• For an unsettling read on Equifax’s history and snooping policies you can start on page 91 the Equifax lawsuit complaint. Here is an excerpt:

One of the more infamous data brokers, Equifax, began business in 1899. By the middle of the 20^th century the company “collected the most intimate details of an individual’s life, including information about their race and sexual habits, their church attendance, their home environment, and whether or not they were experiencing marital discord.” This information was often collected from old-fashion snooping, interviewing neighbors, or digging thourgh someone’s trash.

For a nation that carries world-class spying devices in its pocket 24/7, the reality is that there is no such thing as privacy anymore. That doesn’t mean citizens can’t claw back their lives. California has some of the best privacy laws in the nation, yet companies consistently ignore or make it impossible for people to delete their.

• The Markup “Caught companies making it harder to delete your personal data online” by using secret code to hide delete requests from Google Search. A U.S. Senator is now pressing these companies for answers.

• Russia, bypassing private sector surveillance, is ordering its state-run messaging app MAX to be installed on all cellphones, denying it will use the application for spying or suppressing dissent. (Reuters)

• For a lighter data-breach story, read about the Panama playlists, a “hack” that exposed non-private playlists of around 50 prominent celebrities, politicians, and journalists. Even though all the information was already public (much like the data brokers collect), the author’s of this story point out it is still a little creepy having the information “exposed.”

Well into the social media age, we have grown accustomed to curating ourselves online. Picking top-four movies on Letterboxd is an intentional way to showcase our taste. We decide which vacation photos are most flattering or funny to post to our permanent grid on Instagram. Those books on a shelf in the Zoom background are chosen with care.

But these Spotify playlists were different, created for enjoyment, not for display. It was like seeing someone’s Netflix watch history — a slight invasion and a chance to judge what someone actually consumes rather than what the person claims to like. (NYTimes, Gift Article)

Back to school

17 states and the District of Columbia [are] starting this school year with new restrictions [on student’s phone use], bringing the total to 35 states with laws or rules limiting phones and other electronic devices in school. This change has come remarkably quickly: Florida became the first state to pass such a law in 2023. (AP News).

South Korea, France, Finland, Italy, the Netherlands, and China all have some sort of cellphone ban as well. (NYTimes)

Cellphone bans have mixed reviews; students seem 50/50, teachers love them, and parents generally don’t. While there aren’t definitive studies that can show benefits of banning the phones yet, anecdotal results are pretty straightforward: there is one less disruption in the classroom.

Who is the main stakeholder left out of the cellphone ban story? Social media companies—which are arguably the rootcause for the distractions. Meta has been busy, however. The Tech Transparency Project’s latest investigation found tactics that follow closely to Big Tobaccos tactics a generation ago:

Meta uses an array of influence tactics to try to shape the public narrative around kids and social media. [It] funds a collection of parent and child safety groups, including the National PTA, that go to bat for its initiatives involving kids.

Meta has also created something called the Trust, Transparency & Control Labs that publishes reports in support of its kid-focused products. Meta has at times framed these “labs” as a separate organization to regulators and others. [It also] has funded an array of academic research projects that foster a more benign view of Instagram, helping to support the company’s contention that academic research is inconclusive on the topic of social media’s impact.

Generative AI and cheating are another major concern for schools. How can you test when your students (even the good ones) are going to cheat? The answer is very simple: oral exams and blue books. No one likes either, but they are both tried, trusted, and proven ways to test a student’s ability to retain and analyze material. This guest essay discusses some of the history surrounding oral exams and advocates for their return, as well as the use of handwritten in-class essays. (NYTimes. Gift Article).

Foreign influence

The citizen is the bedrock of a republic, and their vote is sacrosanct. Interference, especially foreign interference, is a horrific perversion of democracy. Two stories this week:

How China Influences Elections in America’s Biggest City. The Chinese Consulate in Manhattan has mobilized community groups to defeat candidates who don’t fall in line with the authoritarian state. (NYTimes)

In New York City, social clubs backed by China undermined a congressional candidate who once challenged the regime on Chinese television.

They helped unseat a state senator for attending a banquet with the president of Taiwan.

And they condemned a City Council candidate on social media for supporting Hong Kong democracy.

In the past few years, these organizations have quietly foiled the careers of politicians who opposed China’s authoritarian government while backing others who supported policies of the country’s ruling Communist Party. The groups, many of them tax-exempt nonprofits, have allowed America’s most formidable adversary to influence elections in the country’s largest city, The New York Times found.

Algorithmic Foreign Influence: Rethinking Sovereignty in the Age of AI. Code now governs what users see, say, and know—across borders, without consent. It’s time to rethink what foreign influence really means. (Lawfare)

Recommender systems, large language models, and machine translation tools now shape civic discourse around the world. They promote certain narratives, erase others, and define what information is available to users—often in ways that reinforce inequality or favor dominant voices. Crucially, they do this without any intention to interfere. They act through infrastructure, not ideology.

This raises a difficult question: If artificial intelligence (AI) can reshape a nation’s public sphere without direction from a foreign power, is it foreign interference?

At first glance, the answer seems obvious: Interference requires intent. Under international law, the principle of non‑intervention is grounded in the assumption that harmful acts are purposeful and attributable to a state. Algorithms are neither. They lack agency, identity, and motive.

But if the outcomes—distorted political discourse, marginalized languages, eroded cultural autonomy—are functionally equivalent to classic interference, shouldn’t the law take effect to address it as such?

Breakneck

Dan Wang’s book, Breakneck, compares the state of engineering in the United States and China and has garnered rave reviews. It is on my reading list for as soon as I return to the states,, and I intend to write a book review of my own as soon as I finish reading it. In the meantime, here are two interviews with the author and two links where you can buy the book yourself. (Amazon. Bookshop.org)1

Rest of World. Why China has a tech manufacturing advantage over the U.S.

On industrial policy:

I’m really skeptical that Trump’s tariffs are going to boost manufacturing. The U.S. has lost some 40,000 manufacturing jobs since “Liberation Day” in April.

I think that Biden’s industrial policy instincts were more correct, but the execution was poor. It didn’t really seem like they needed to move at breakneck speed in anything aside from semiconductors. In everything else, they were much more obsessed with procedures rather than delivering results.

China Talk. Dan Wang.

On pluralism:

Jordan Schneider: … Someone you met in Chiang Mai who left China told you that contemporary China “feels like a space in which the ceiling keeps getting lower… To stay means that we have to walk around with our heads lowered and our backs hunched.” You also write that, “After six years in China, I missed pluralism. It’s wonderful for me to be in America now, in a society made up of many voices, not only an official register meant to speak over all the rest.”

On political freedoms & good governance:

Dan Wang: We can accept all of these [problems of city living in America as opposed to the efficiency in China]. But what we also have in New York — and this is part of the reason that I’m drawn back to the US again — are bookstores. At these bookstores, one can find books very critical of the US government, very critical of both the Republicans as well as the Democrats, who have both made incredible errors. For the most part, though there have been some restrictions on protests under Trump’s regime in America right now, there still is broad latitude for people to protest all of his illegal or inhumane actions, and that is still very real. Protest culture online is also very real.

What I’m saying is that I hope we don’t have to choose. The United States should be able to have, at its present levels of tolerance of dissent, as well as very functional cities that have good subways, good bus systems, nice airports, and where people are able to get around and have a rate of improvement that doesn’t come at truly absurd financial costs.

Intel-ligent stakeholders

What do President Trump and the country’s leading socialist Senator Bernie Sanders have in common? No, it is not that their net-worth is well above 7-figures or that they own multiple houses, it is that they both support the government taking shares in private corporations.

I’m ambivalent about the government’s move to take a stake in Intel, especially without Congressional Approval, but there are some potential upsides to it as well. This Lawfare article provides a good history of U.S. Government takeovers and the legality of the Intel Stake. The author also has a good LinkedIn post on the subject:

[…]

4. According to Intel's SEC filings on the government investment, in exchange for 9.9% of Intel, the government (a) agreed to disburse most of the remaining CHIPS money (about $6 billion from Commerce, and part of Intel's DoD grant), AND ALSO agreed to deem Intel as having complied with its grant obligations, e.g., to have completed most of the fabs. This despite the fact that several of the fabs are not, in fact compete. In exchange, Intel gave the government a 9.9% stake. (Intel is still on the hook for some of its DoD-specific obligations)

5. Basically, instead of CHIPS Act money being tied to Intel actually building all the fabs, the money is now tied to the government buying into Intel. That is a different thing. The specific deal is also different from the concept of the government getting shares as part of a grant to build fabs. In the SEC filing, there is less emphasis on fab building and more emphasis on government ownership.

In other news

• Here is a tracker on the civil service firings (most of which are from the DoD purge).

• And here is a story about a pediatric brain cancer treatment facility that is losing funding.

• For an audio recording to be admitted as evidence in court, all that is needed is for someone “familiar with the voice” to vouch for its authenticity—a significant problem given how good AI deepfakes are. This Lawfare article lays out the problem and proposes some policy solutions.

• Where does your e-waste go? Most of it goes to India. If you “recycle” old electronics, you should take a look at this story and see where they end up and who sorts through them.

• Two new reports are out, one on Ukraine and the Law of War, and the other from SIPRI on bias in military AI.

Before you go

An excellent Guest Essay in the New York Times from Eric Schmidt and Selina Xu is titled “Silicon Valley Is Drifting Out of Touch With the Rest of America.” They discuss how Silicon Valley’s obsession with AGI, as opposed to integrating AI with the current economy—something China is currently excelling at—is damaging and not what Americans want.

The essay is also fitting with a Financial Times story titled “San Francisco without the rich: a photo-essay,” which captures what San Francisco was like when it was a high-functioning, romantic northern California city (FT Free Story).

It was startling to learn how the Nazis almost successfully covered up the killing of 1.9 million people. Most of the images we have of Auschwitz came from pictures taken by German soldiers in their ever-banal and meticulous documentation process. The SS tried to destroy much of the evidence of genocide once they realized the Soviets were advancing toward their position. These images, by chance, were not destroyed, and, also by chance, were found by a young girl who survived the camp.

Surviving eyewitnesses were nearly nonexistent. The Sonderkommando, Jewish prisoners forced to dispose of the victims’ bodies and therefore some of the few non-Nazi eyewitnesses to the crimes, barely survived the camp—only 20 of the thousands of Sonderkommando who were in Auschwitz survived and then testified about the crimes.

The evidence collected at Auschwitz and other camps was crucial to convicting the war criminals at Nuremberg and other subsequent trials. But if a crime that large could be nearly covered up 80 years ago, could that happen today? Sensors are everywhere and documenting everything, but just because wars are being live-streamed doesn’t mean the world knows what is going on 100% of the time. Moreover, there is a difference between documenting a crime, using that documentation in court, and then believing the documentation is real.

I’ve compiled some articles that discuss what technology and systems make documenting war crimes easier, where the challenges are, and what technology is making proving crimes more difficult (hint, it’s deepfakes).

What is easier

In some ways, documenting and proving crimes has never been easier. Sensors are everywhere. Anyone using Google Maps can access pretty good and up-to-date satellite imagery. NASA operates a site that has even better imagery available for free. For a couple of hundred dollars, you can get even better imagery from a commercial provider, such as SkyFi. Larger enterprises with more capital can have capabilities previously reserved for only the most sophisticated state actors.

One of the best-known names in Open Source Intelligence (OSINT) is Bellingcat.

They conduct investigations, provide guides, resources, and case studies, and actively help archive online evidence. Their OSINT toolkit offers tools for investigating corporations, identifying phone number leaks, following key individuals on X for OSINT news, and more. For the amateur sleuth with no experience, the world really is at their fingertips.

Despite all the OSINT tools out there, it really does not get easier when war criminals live-stream their war crimes.

It doesn’t take much digging on X to find accounts that post combat footage. Some of this footage is valuable OSINT information from trusted sources. Other times, it is simply war porn. The Wagner Group is notorious for doing this, and they are monetizing their war crimes without consequence. While live-streaming war crimes certainly falls into the “making it easier to prove” category, as I’ll discuss later, you still have to bring the criminals to trial. I encourage everyone to read the linked story about the Wagner Group. It shows the disparity between ICC investigations around the world and rightly draws parallels (or a line of continuity) from the Heart of Darkness and King Leopold’s Congo.

We all carry surveillance devices around 24/7.

Everyone carries a small, portable device equipped with a microphone, cameras, and GPS systems constantly pings your information to and from cell towers and wifi stations while updating scores of applications (some of which you cannot see) with various pieces of information, including your location, what you are buying, how you are buying it, what you are watching, and who and when you are talking to someone. This privacy nightmare is a criminal detective's wildest dream. Here are two stories where a criminal’s cellphone was enough to land a conviction.

• 3 Teens Almost Got Away With Murder. Then Police Found Their Google Searches

• Both sides use trove of cell data at Alex Murdaugh trial

The challenges

With a flood of new technology, there comes a flood of new challenges. When there are massive amounts of digital data, that data needs to be stored securely, authenticated, and be admissible into a court of law. Here are some stories detailing precisely those challenges.

The Problems of Digital Evidence: How authenticating, securing, and protecting digital evidence is problematic at the world's International Criminal Court

While new technologies and new investigative methods hold great promise for improving avenues for international justice in many ways, authenticating digital evidence is a major challenge. Authentication ensures digital information is what it purports to be—in other words, it has not been altered, manipulated or forged. Authentication is important, as digital evidence is more susceptible to compromise, forgery, manipulation, and deletion than traditional physical, documentary, or testimonial evidence. Unlike other forms of evidence, there is no analog reference for most forms of digital evidence. In addition, social media’s amplification of disinformation in international conflicts has increased the ‘fog of war’, running the risk of entrenching potentially misleading or false information in eyewitnesses, investigators, and judges.

What the Afghanistan Withdrawal Teaches Us About Safeguarding Human Rights Evidence

In [2021] mid-August, the Taliban swiftly advanced to take control over the Afghan territory and government. […] While vulnerable individuals were understandably and rightfully prioritized in the evacuation efforts, non-human materials and resources were left behind in the chaos of the Taliban’s takeover. Included in the materials left behind were hardcopy archives that international civil organizations and the Afghan Independent Human Rights Commission had collected to document human rights abuses committed by the Taliban, Afghan forces, and other actors. Over decades, human rights investigators had amassed thousands of pages of evidence including victim affidavits, photographs and videos of human rights abuses, and forensic evidence related to mass graves. Sensitive personal identifying information of victims, witnesses, and human rights documenters was also embedded in these documents. […]

But it did not have to be this way. Modern storage and verification technologies could have been leveraged to neutralize these risks. This article delves into the technical details of attempts to modernize today’s evidence safehouses (including those of the Starling Lab, with which this author is affiliated) and how these technologies’ might be deployed in future investigations that take place in high-risk conflict zones.

While a citizen, journalist, and lawyer may all deem something an “atrocity” if it is actually a “war crime” (and one that has a chance of being successfully prosecuted), it takes a trained eye to distinguish it. These trained eyes are few in comparison to the massive amounts of data that need to be sorted. AI is partly solving this problem, but the gap between identifying something as a war crime and prosecuting it remains significant. This story describes just that journey of an ad hoc group of Ukrainian citizens who went to work during the opening hours of the full-scale invasion:

The Hour These Hostilities Began’: Ukrainians Mobilize to Document War Crimes

Investigators must track down and verify the sources of such information, not an easy task when drawing from the Internet, where the habit of posting under a pseudonym and the proliferation of “deep fakes” can complicate the authentication of data. Villagers coming to retrieve, honor, and bury their dead, and the journalists rushing to capture and report the ravages of Russian aggression, are engaged in understandable, important, and humane endeavors. But their early presence at a crime scene can compromise the ability to establish identifying information about the attackers that is critical in building cases. And all this technological progress can be hindered by the slow pace of traditional judicial procedures.

AI-Generated Voice Evidence Poses Dangers in Court

Evidence rules are also outdated. This piece discusses the antiquated rules around allowing a voice recording into evidence. All that is needed is someone to be “familiar with the voice” to allow the recording. You can imagine, with AI voice fakes being so good, plenty of fake recordings may be admissible as legitimate evidence unless the rules are changed.

The difficulties

Deepfakes. Deepfakes are a massive problem. For starters, they create problems where they don’t exist. This Guest Essay by Senator Amy Klobuchar, a recent victim of a deepfake, articulates the problem.

Amy Klobuchar: What I Didn’t Say About Sydney Sweeney

As anyone would, I wanted the video taken down or at least labeled “digitally altered content.” It was using my likeness to stoke controversy where it did not exist. It had me saying vile things. And while I would like to think that most people would be able to recognize it as fake, some clearly thought it was real. Studies have shown that people who see this type of content develop lasting negative views of the person in the video, even when they know it is fake.

Creating controversies that do not exist applies to war crimes investigations as readily as it does to politics. As mentioned above, investigative resources for international crimes are scarce, and having investigators chase crimes that didn’t happen is an unnecessary distraction—and possibly a future tactic for states and individuals trying to avoid punishment. Even worse, the stigma that comes with even being accused of a crime is high, and false accusations can have long-lasting negative repercussions.

Deepfakes present a new type of deniability.

If everything could be fake, then something real could easily be waved off as such. Here is an older post I wrote about just this.

Weekly Roundup: Deepfakes Meet Swift Resistance: Politicians and T-Swift confront deepfakes and the war on truth

In sectors like [politics] and journalism, this “new kind of deniability” means that organizations or individuals must go to greater lengths to prove facts than they ever have before. If every scandal, unsavory piece of reporting, or inconvenient truth could plausibly be a deepfake, then the threshold for truth will be extraordinarily high. This could manifest in a way where everything reported is false until proven beyond a reasonable doubt to be unimpeachably true. Or, it could manifest in an extreme version of “he-said, she-said,” where the truth-teller is simply the person with the more amenable personality.

Lastly, despite all the new tools, digital evidence, and the seemingly thousands of eyeballs watching everything all the time, there are still massive amounts of unknowns.

Here is another excerpt of the “Hour the Hostilities Began” story linked above.

All of these matters pertain to the slaughter of civilians. We also confront a whole different category of offense: the forced relocation of thousands of Ukrainian citizens, including children, to Russia. Their location is unknown, communication is impossible, and investigators have no access to Russian terrain. Proving the crime of kidnapping will be hard. And establishing factual links between so many individual atrocities and the criminal intent of Russia’s political and military leadership will be harder still.

What Happens When There is no Nuremberg?

One of the reasons the Holocaust has proved such an enduring testament to the perils of authoritarianism, racism, the banality of evil, and aggressive war is that victorious nations put the crimes on trial, publicly. Evidence was collected, presented, and tried. The accused had a chance to mount a defense, and then, many of them, because of overwhelming evidence, were sentenced.

Except for fringe groups, very few doubt the existence of the Holocaust. Perhaps, a close counterexample to what happens when there is no public trial or documentation of evidence is the Genocide of Poles in Volhynia, and subsequent forced resettlement of Ukrainians a few years later. Volhynia is still talked about among the Polish population with great resentment and anger. Many in the rural Polish borderlands hate the Ukrainians as a collective for Volhynia. Perhaps if the perpetrators were publicly held responsible, there would have been no need to force 150,000 people from their homes in revenge, and there would be some closure for the descendants of the victims.

Ukraine is not the only nation falling victim to atrocious war crimes. China is actively committing a genocide against the Muslim Uyghurs. The famine in Gaza is a war crime. Sudan may be in the midst of a genocide, and is undoubtedly in the midst of an oft-overlooked famine. For an undeniable record of these atrocities to survive, there must be accountability for the criminals. Accountability can only be achieved through public trials and the presentation of meticulously documented evidence. Public trials can only occur if a nation has achieved victory over the criminals. The world narrowly avoided a cover-up and Auschwitz, and despite the challenges with modern digital evidence collection, there is no excuse for war crimes to go unprosecuted in the 21st century.

Resources

Below are some resources you might find useful for OSINT, Privacy, and the ICC.

• The Inner Workings of the International Criminal Court

• Berkeley Protocol on Digital Open Source Investigations: A Practical Guide on the Effective Use of Digital Open Source Information in Investigating Violations of International Criminal, Human Rights, and Humanitarian Law

• Bellingcat’s Online Open Source Investigation Toolkit

• Pioneering Oversight’s Privacy Primer

Other Pioneering Oversight stories on War and International Law

(Parts II and III of “Regulating Autonomous Weapons” are linked in Part I)

If you’ve enjoyed this article, please consider supporting this publication in a few easy ways.

• Like and share this article or Pioneering Oversight with a friend.

Share Pioneering Oversight

Share

• If you aren’t subscribed, consider liking and sharing the publication anyway, and of course, subscribing.

Subscribe now

• If you like what I’m writing, I always appreciate restacks, quotes, and comments!

But it is an important issue. What is happening is the executive branch of the U.S. government is exercising control over individual companies in a way I believe is unprecedented, at least post-WWII. I’ll try to give as brief a background of the H20 chip wars as I can and then explain what these policies mean for the future of technology regulation, national security, and American capitalism.

What exactly is happening?

Here is an excerpt from a ChinaTalk podcast with , , and an old professor of mine, . The whole podcast is worth a listen/read.

Jordan Schneider: Lennart, what is the H20? Why should people care about it? What were the first few months of the Trump administration doing when it came to this chip?

Lennart Heim: The H20 is a chip that NVIDIA designed as a response to export controls in 2023. It’s the typical game: you draw some lines, and then new chips get created right below those lines. The H20 is exactly such an example, but it did a neat trick.

It maxed out the specifications that are not controlled — memory bandwidth. They put the best high-bandwidth memory the world currently has on this chip and created an export control-compliant chip that was introduced at the beginning of 2024, a couple of months after the updated controls. The chip was sold throughout 2024 with lots of interest.

When the Trump administration started in January, the Biden administration didn’t get around to addressing this problem. Many officials spoke out in favor of taking action, but they never got to banning it because of many stakeholders, different opinions, and running out of time.

Trump then banned this chip, as reported in April 2025. Not through the normal regulatory process, but by using a tool called “is-informed” letters, which are pretty fast. You can send a letter to the companies that produce these chips telling them they can’t sell these chips anymore because you suspect an export control violation is going on. The administration argued this chip was simply too good.

So, after the April 2025 ban, what happened? Well, in July 2025, President Trump then decided the chips could be sold to China with a 15% “export tax”, but,

“Beijing appeared to hesitate. Trump’s reversal looked like a straightforward win for China. Yet Beijing ordered its regulators to investigate whether Nvidia’s chips have “loopholes and back doors”. Now that Nvidia has agreed to pay 15 per cent of its Chinese H20 chip sale revenues to the US government, it can restart sales — but first it must convince Beijing that its chips do not pose a security risk.” (FT Gift Article, FT Subscribers)

Are these chips a national security problem?

What are the pros and cons of the ban? What are the national security implications for both countries? Again, from ChinaTalk:

Lennart Heim: From my personal point of view, banning the chip was a big success. This chip should not be sold. We need to reduce our thresholds — this is simply too good of a chip. […]

There are arguments against selling AI chips because selling helps upgrade the Chinese AI ecosystem that’s going to compete with America’s. There are specific applications of the chips that we would be selling to China that we would be very uncomfortable with — military ones, intelligence ones, or broad human rights violations that you wouldn’t want American technology to be helping to further.

There are also arguments in favor of selling. These include the idea that selling NVIDIA chips would retard domestic chip development, making it harder for SMIC and Huawei, and whoever else wants to try to build domestic AI chips to find a marketplace. There’s also the idea that selling chips into China would maintain Chinese dependency on the US stack, keeping Chinese developers using CUDA, building infrastructure around US technology. There’s some broad soft power and agenda-setting advantage that China's use of NVIDIA hardware will give to the US going forward.

Chris Miller: …The question is, how best do you get [to lead in AI]? One argument is that you restrict compute access and thereby hobble the growth of Chinese AI firms. A second argument is that you try to, as Secretary Lutnick has said, get China addicted to the AI stack. The question to ask is: how addicted are they willing to become? How addicted could you make them? Can you leverage that addiction in the future, or not? These are where the empirical questions are focused.

America isn’t the only player in this game; China gets a vote, too. “China’s been racing toward self-sufficiency,” […] “This is a strategic goal of the Chinese Communist Party that has persisted now across four different US presidencies.” (Bloomberg)

Can you actually stop the export?

Controlling the export of goods is difficult. The process basically works like this: If a company wants to export a controlled product (not military related), it probably falls under the jurisdiction of the Commerce Department’s Bureau of Industry & Security (BIS). A company would apply for a license, seek permission, and either receive it or not. If a company fails to obtain the requisite permission, the BIS has sleuths looking for clues that a company is exporting without permission, and there are civil and criminal penalties.

Enforcement is a difficult job, especially when companies either lie about or are unaware of the end user. This is why the U.S. is embedding trackers in shipments. It may seem like an old-school method, but a lot of what BIS is old-school detective work.

Shifting sands of U.S. capitalism?

The United States, particularly the executive branch, is increasingly controlling companies through threats made by the President and by owning actual stakes as an investor.

Recent examples include President Trump’s demand that Intel’s chief executive resign; the 15% of certain chip sales to China that Nvidia and Advanced Micro Devices will share with Washington; the “golden share” Washington will get in U.S. Steel as a condition of Nippon Steel’s takeover; and the $1.5 trillion of promised investment from trading partners Trump plans to personally direct. (WSJ) [The Pentagon is also taking a stake in a rare-earth mining company.]

I’m not sure how I feel about these stakes. There is a large part of me that feels it is probably good. I think 99% of companies, when push comes to shove, would support the United States in national security issues. Microsoft has long helped the government in areas like cybersecurity. The Russian exit of Western firms (minus Pepsi) at the start of the full-scale invasion was impressive. But that 1%, and unfortunately, right now it aligns with firms that are highly related to national security, I think, would sell America out on a dime. I wrote about this a few months back:

On the other hand, this is undemocratic. The state taking shares and controlling interest in an industry crucial to national security is one thing, but having those decisions not come from Congress is problematic.

Corruption characterized by preferential treatment is the primary concern. Societies with top-down executive branches and weak legislative systems, coupled with courts that are undermined or lack legitimacy, are ripe grounds for corruption. What will the oversight mechanism be for a President who can unilaterally control corporations by controlling board members’ appointments or as a shareholder, while simultaneously being responsible for enforcing laws against the very corporations they control? Case-in-point:

Trump Has Dropped a Third of All Government Investigations Into Big Tech

According to a new report from Public Citizen—a nonprofit government watchdog—the Trump administration has dropped one third of all pending enforcement actions against tech companies. Those same companies collectively spent $1.2 billion on political contributions since 2024, most of it going to Republicans. Some of it went to Trump directly.

With the executive branch having more and more leeway in controlling the economy, such as potentially using the export tax to greater effect across other industries, America has abandoned free market capitalism. While this issue may make many radicals on the right and the left happy, it should give alarm to those who do not want the government to have absolute control.

Subscribe now

Should the U.S. adopt the China Model?

What if the current administration’s policies work? Is it worth giving up individual freedom, both politically and economically? I think there is a strong minority of people in this country who would say yes. I think there is a strong majority of people who are happy that someone is “getting stuff done” and may not see the day-to-day value in these freedoms, especially if their food and housing costs are so expensive. Below is an excerpt from ’s Con-Current. I recommend reading the whole piece. It compares the United States and China in terms of their optimism and pessimism—liberal reformers should take notice of the consequences of their inaction.

Afra: China's "indefinite and pessimistic sentiment" often confuses me and I've been contemplating its roots. When visiting China, you experience impressive infrastructure and consumption options that are noticeably more abundant, higher quality, and futuristic than those in Europe and America. The surface prosperity is undeniable. Each time I take the high-speed train, I'm genuinely impressed and never take it for granted. Given China's remarkable achievements, why does EVERYONE I meet continue to harbor such profound pessimism? One possible explanation is: you never know when the iron fist (铁拳) from the state will come down. This feeling that it could strike at any time creates constant tension, much like playing Russian roulette.

Du Lei: What you said is exactly my reaction when thinking about this. I'm trying to find a more tactful, gentler way to express this viewpoint. I think certainty about the future might be the key OS difference between the two sides.

America also faces uncertainty, but its basic governance logic is very bottom-up. In a strongly bottom-up system, every participant has greater agency. With stronger agency, people think: "Even though I'm uncertain about the future, I can remain optimistic because I always have multiple chances. If I don't win this round, I can try again." This mindset exists because the entire system operates as a bottom-up free market approach: I can always have another opportunity.

But in a very top-down model like China, success or failure often doesn't depend on individual actions, but on which side you stand. Top-down decisions directly determine your fate in this game played with a visible hand. In this situation, all participants lack agency. People think: "As a whole, China will definitely win, but as an individual, I'm not sure if I'll be the tears of the era or a sacrifice." In China, everyone believes the state will ultimately succeed, but no one knows whether they'll be the victor or the price paid for victory. This top-down versus bottom-up approach shapes how individuals judge their situation. People aren't pessimistic about collective success, but are more pessimistic about their personal outcomes.

In other news

• wrote a great piece on technology and women’s reproductive rights at Reboot. Eliza brings up many interesting aspects of the argument on what principles should guide the use of these new technologies.

Reboot

📏 repro/acc?

Kernel Magazine Issue 5 is out! We’re sharing another piece from the issue — Eliza Steffen’s look into the strange coalitions and new political alignments in the evolving world of reproductive tech. If you were at the Kernel Launch in SF, you got a sneak preview of this piece! For the rest of you…

Read more

8 months ago · 14 likes · Eliza Steffen

• Last week, I wrote a long piece on how technology is making the differences between civilian and military targets moot. It’s a long read, but important for anyone interested in national security, privacy, and local governance. The sensors that create this technological civilian-military entwinement are ones we use every day, and it is going to be state and municipal level governments and organizations that feel the effects.

• A federal judge struck down a CA law that prevents deepfakes during election cycles. The law was struck down not because of 1^st amendment issues, but because of Section 230. Deepfakes may or may not have valid First Amendment protections, and those protections may or may not need to be limited during election cycles. However, the fact that we cannot even hear the merits of these arguments due to Section 230 liability is obscene, and this is another reason why that protection needs to be narrowed or gutted.

• The United Kingdom implemented an age verification law, causing traffic to porn sites to drop by nearly 50%. I’ve been supportive of these laws in the past while acknowledging there are dire implementation consequences that need to be monitored closely to ensure governments don’t restrict other content. Many opponents of these laws argue that they are technically infeasible to implement, and users will circumvent them with VPNs. The UK law will be a great place to test this hypothesis—we will see if porn website traffic increases back to normal rates in the future.

• This is a great Lawfare article detailing the need to reauthorize the Cybersecurity Information Sharing Act of 2015.

• Bellingcat has updated its Open Source Archive tool, which preserves online evidence and aids organizations in investigations. This tool may be of interest to anyone fighting back against groups or governments that “cancel” history, facts, figures, or engage in revisionism.

If you’ve enjoyed this article, please consider supporting this publication in a few easy ways.

• If you are already subscribed, please like and share this article or Pioneering Oversight with a friend.

Share Pioneering Oversight

Share

• If you aren’t subscribed, consider liking and sharing the publication anyway, and of course, subscribing.

  Subscribe now

• If you are a big Substack fan and you like what I’m writing, I always appreciate restacks, quotes, and comments!

  ---

What is different today and will be more so in upcoming conflicts is that civilians are inextricably entwined with military sensors, technologies, and a manufacturing process that derives from commercial-off-the-shelf products. This entwinement, when coupled with the increasing need for societies to engage in “total war” economies and mobilization, presents a sordid proposition: everything is a military target.

When democracies go to war

The norm is that democracies are surprised at the start of all wars. The only way to respond and fight through the shock of surprise is to have a broad range of military capabilities to provide many options to government and enhance military resilience. Only a balanced force can do this. (, Futura Doctrina)

Ukraine faces a similar dilemma today that the United States faced during the Global War on Terror: personnel shortages. Both are also coping with it the same way. Ukraine has revolutionized military tactics and technologies with drone warfare, a strategy of attrition meant to counter Russia’s superior manpower advantage, while preserving the lives of as many Ukrainians as possible. In the war against ISIS, the United States made a political decision to use technology, artillery, and special operators at high rates that had severe consequences, shielding the American electorate from the consequences of war. I highly recommend the linked NYTimes article to understand the full consequences of using a small military footprint (New York Times Gift article link).

The ability to prosecute a war like the U.S. did during the Global War on Terror (GWOT), characterized by technological solutions meant to keep American soldiers off the ground, is probably over for the time being. The Futura Doctrina piece linked above discusses the balance militaries will have to take in future wars, including dramatically increasing their armed forces through rapid mobilization.

The experiences of Ukraine and Israel in the past three years have demonstrated the need for rapidly expandable military forces. While the core professionalised force has played the central role in American and Australian forces since the end of the Vietnam War, this model is increasingly not fit for its strategic purpose. The vast array of challenges that military forces must now respond to – at home and abroad – calls for larger forces. At the same time, an array of new skillsets – from drone pilots to AI experts to the need for more logisticians to support integrated global operations – are needed in military organisations used to contracting these capabilities. Finally, the ability to recruit sufficient personnel has come under increasing pressure in societies where there is declining trust in institutions and many people are reticent to commit to long-term careers in a single organization.

This idea was echoed by Jacquelyn Schneider and Julia Macdonald in their recent piece “How to Lose the Drone War: American Military Doctrine Is Stifling Innovation” in partial response to Hegseth’s “unleashing drone dominance” memo.

[The U.S.] cannot simply copy the Ukrainian (or Israeli) drone strategies. Before rushing toward procurement, U.S. defense strategists need to articulate a new theory of victory, reviewing the beliefs and assumptions that undergirded the last 50 years of technological acquisitions.

The United States’ delay in adopting the new generation of drones is a product of strong convictions that it acquired over the past 60 years of fighting wars: that it could, and should, leverage remotely operated technology to win quick conflicts fought from a distance. The United States believed that it could rely on relatively expensive drone technology to save pilots’ lives, deliver real-time intelligence straight to decision-makers, and enable precision targeting.

Before rushing to invest in a fresh wave of technologies, however, U.S. defense planners will need to review fundamental beliefs and assumptions that guided their acquisitions over the last half century. They will have to reconsider the American public’s tolerance for casualties, reevaluate long-standing procurement processes, and wrestle with the different services’ tendency to push for bigger, pricier systems.

For half a century, the United States built a military based on the belief that the American public would not sacrifice its blood but would be willing to spend its treasure. As the deficit balloons and the U.S. electorate shows an increased appetite to punish presidents for inflation and wasteful government programs, U.S. leaders can no longer simply assume they can throw money at expensive technologies to mitigate their own political risk.

One upshot from these pieces is that jobs that used to be done by the military and have since been contracted out will probably have to come back to the military. Not every American will have to be a soldier, but millions more will have to don the uniform for the next war to be successful.

Everyone is a drone maker

“Without reevaluating the American way of war, no amount of new drones will be able to defend the United States against wars it doesn’t want to fight.” (Foreign Affairs)

Ukraine has a decentralized military-industrial complex. Everything from R&D to testing to procurement happens mostly around small and medium-sized businesses and at the unit level. This is not to say the entire system from start to finish is decentralized. Ukraine still has a centralized procurement certification process—the money to buy supplies still mostly comes from the national government, although cities, towns, and NGOs also supply capital and equipment.

Many small businesses within a particular vertical, such as demining technologies or FPV development, lean on each other for co-development and co-procurement. That centralized procurement certification procedure is difficult and takes time, so many small businesses will band together to sell “one product” to the government instead of each company going through the process. They will license technology to each other, formally and informally, to create a singular product. One company might make cameras, another solenoids for motors. The end product of this consortium is a usable drone that can navigate the procurement process as a single product.

Knowledge is also dispersed. One contact of mine told me he was impressed by how non-military, non-engineering, “every-day” Ukrainians know how to assemble drones. There are schoolhouses throughout the country that teach the skill. Pop-up workshops are available. A Ukrainian 20-something told me on the train that he hosts drone-making parties.

These consortia of knowledge sharing, either in the procurement process or on the training side, are just one example of the integration of the population into the war effort. This decentralized “total war” system lends itself well to a democracy fighting back an invasion. However, as I’ll discuss later, the dissolution of any clear distinction between what is civilian and what is military has consequences.

Two more reads on drone warfare that can be read as a point-counterpoint:

1. Ukrainian Drone Battalion Forms Fire Wall Against Russian Advance. (Kyiv Post)

2. From the Frontline: I’d Rather Have a Mortar Than a Drone. (CEPA)

Internet problems

It is not just that civilians are choosing to make drones to support their soldiers that creates this technology-driven civil-military entwinement. The sensors and communication methods that are vital to civilians' livelihood and used every day are not just nice-to-have “dual-use” items to the military, but are also vital to military combat operations.

The limiting factor in military operations can nearly always be attributed to communications in some way. Drones are limited by the electronic warfare environment in which they fly. Orders are limited by how quickly they can be disseminated. Requests for fire support will only be fulfilled if the artillery battery receives them.

A new report confirmed suspicions that Elon Musk personally ordered the shutting off of Starlink during a major counteroffensive in 2022.

The move reportedly caused front-line units to lose connectivity, disabling drones and disrupting artillery coordination. A Ukrainian military source told Reuters that the outage directly contributed to the failure of the encirclement operation. (Kyiv Independent)

While Musk denies this happened, the unambiguous part of the story is that a lack of connectivity hampered active military operations. Soldiers more than likely lost their lives because of this.

The communication issue cuts both ways. Russians heavily rely on mobile internet for their day-to-day lives, more so than other nations, as it is the primary source of connectivity for most Russians. Not only is it their only source of connectivity, but Russians have to use their phones to engage in all sorts of day-to-day transactions that are not possible in an analog form anymore. They cannot effectively buy food, gas, pay rent, or much of anything without the internet. Ukraine has been taking advantage of this by using Russian networks to navigate during their deep strikes into Russia. (The New York Times)

The Russian response to this problem is to shut the internet off for everyone. It is unclear if this is having any effect on Ukraine’s ability to navigate its drones inside Russian territory, but it is affecting the Russian population. The reality is that the ubiquity of sensors, communication networks, and digital infrastructure is dual-use, and militaries, especially an invading one operating from over the horizon, are going to need to take advantage of these sensors. Many of them (think of the millions of ring cameras) may not be able to be shut off like mobile network towers. And if they are, there are going to be consequences for a population that is actively trying to resist the invaders at the same time. The consequences will be much worse for the civilian population using them if the military wanting to shut them off is not friendly. These sensors provide direct “effective military contributions,” and, as Russia has shown, it doesn’t matter if they are on top of or amid civilian infrastructure, or if civilians rely on them for their livelihood.

Two other reads on communications and war, and a primer on quantum technology:

1. Hackers Disrupt Russia's Drone Weaponization Network. “A volunteer-run network of service centers halts custom firmware updates for DJI drones following a cyber attack.” (Oleg Shakirov @fromcyberia Packets from Cyberia)

2. To keep the world’s data flowing, countries need to quickly fix broken undersea cables. This is an excellent history of undersea cable tampering. (The Bulletin)

3. Military and Security Dimensions of Quantum Technologies Primer (SIPRI)

Cyber-enabled terrorism

John Steinbeck’s The Moon is Down is a great story about a small town resisting an occupying force. It is a short read, and I highly recommend it. At one point in the story, the town’s residents discuss trying to get allied nations to drop sabotage supplies from cargo planes—a method that is not exactly inconspicuous.

Airdropped cargo for sabotage, resistance, or not, is not required anymore. Russia is using a “gig model” version of acquiring assets and sabotaging countries allied with Ukraine. This free-lance, Uber for sabotage system has been attributed to the Warsaw shopping mall fire last year, a false-flag operation in Germany during an election cycle, and arson at the U.K. Prime Minister’s home. (Politico)

Inside Ukraine, and to a lesser extent in other countries, Russia is using Telegram to trick and entice people into being suicide bombers. Children and older women have been given packages under false pretenses and delivered the packages to the target while being remotely detonated by the Russian operative. Sometimes the Russians use crypto payments to entice people, payments that usually don’t come. (CEPA)

One key issue behind these attacks is the communication system: Telegram. 70% of Ukrainians rely on Telegram for news, bomb alert information, and government communications. I only use it in Ukraine, and sometimes it is the only way to communicate.

It is probably impossible at this point for Ukraine to disentangle itself from Telegram, and it goes to Mick Ryan’s earlier point about ensuring countries have “a broad range of capabilities.” The Telegram example should be a pressing issue to leaders who do not have to deal with military or national security decisions. It is not conducive to civil resistance to have to communicate with your people using the same app the enemy is using to recruit saboteurs.

In America, X is analogous to Telegram. Think about how often officials tweet out information crucial to the public, or how frequently you read about an official tweeting out information in a news story. If this platform is compromised in any of the ways described above, are there other reliable alternatives?

Another read on this subject:

1. X and WhatsApp Host Arms Traders Tied to Houthis. “Elon Musk’s X and Meta-owned WhatsApp are hosting a thriving arms trade tied to Houthi militants in Yemen…undermining U.S. national security interests” (Tech Transparency Project)

If you’ve enjoyed this article, please consider supporting this publication in a few, easy ways.

• If you are already subscribed, please like and share this article or Pioneering Oversight with a friend.

  Share Pioneering Oversight

  Share

• If you aren’t subscribed, consider liking and sharing the publication anyway, and of course, subscribing.

Subscribe now

• If you are a big Substack fan and you like what I’m writing, I always appreciate restacks, quotes, and comments!

The thin legal shield

In some ways, Ukraine is a poor example of the entwinement of the civilian population with warfighting efforts. It is hard to disentangle your war effort when the war was brought to you by an aggressive authoritarian power trying to eliminate your nation. Russia is not really trying to legitimize striking apartment buildings in international law—it is just blatantly killing civilians. Still, the future of warfare is probably going to require this level of civilian involvement—and certainly this level, if not more, of technological entwinement—whether that nation is being physically invaded or not, and this is going to have severe consequences for the civilian population when the enemy selects targets.

“Targeting civilians and civilian infrastructure, even if it is sometimes used by the military, is illegal under the law of war. The Geneva conventions “prohibit[s] not just useless violence against civilians, but useful violence against them as well.”1

In international law, an object is “either a ‘military objective’ and may be attacked (subject to proportionality and necessity analyses) or it is a ‘civilian object’ and may not be attacked.” There is no such thing, legally, as ‘dual-use.’”

Article 52 in the Geneva Conventions Additional Protocol I says that a:

1. Military Objective “is those objects which by their nature, location, purpose or use make an effective contribution to military action and whose total or partial destruction, capture, or neutralization, in the circumstances ruling at the time, offers a definite military advantage.”

2. Civilian Objective is everything else.

3. Dual-Use Objective is not defined in international law.

Hathaway’s piece (linked here and in the footnotes) presents evidence from states’ militaries that supports the idea that objects labeled as “dual-use” are not causing militaries to be more cautious about targeting them, but are being used as an argument and rationalization as to why the object should be targeted. To counter this trend, Hathaway offers three solutions that could better protect civilians:

1. Militaries must consider the harm caused by the loss of the object through the eyes of the civilian, in addition to things like collateral damage.

2. Objects that do not have a direct connection to the war but are “war-sustaining” should not be targeted.

3. “States should 'incorporate ‘reverberating effects’ and ‘cumulative harm’ into their analysis—civilian casualty rates are not enough to calculate the loss of livelihood.”

I think the first and third recommendations make sense. Regardless of the civilian-military technological entwinement or any “dual-use” discussions, we should be striving, especially in war, to protect civilians in every way possible.

The problem with the second second recommendation is that it assumes that civilians can effectively be separated from the technological entwinement that is characteristic of modern war.

Hathaway’s analysis may lead to an overcategorization of objects that are “war-sustaining” and then be effectively ignored by military planners, as then nothing will be able to be targeted. Alternatively, it may lead to an undercategorization (i.e., ignored completely), which is where we are at with Russia’s war.

Here are some other practical examples of where a “war-sustaining” analysis runs into trouble with civil-military technological entwinement.

• The Commander in Chief of the United States effectively controls the board of Nippon Steel. The Pentagon is taking a 15% stake in MP Materials, a mining company, and guaranteeing a price floor for rare earth minerals the company mines, protecting MP from a potential market flood of the minerals should China resume its exports to the U.S. at full capacity. (WSJ) These rare earth minerals are used for cars as well as fighter jets. The miners and the company are war-sustaining and partially owned by the government. I think most militaries would consider this a reasonable military objective, but it would probably be considered “war-sustaining.”

• Many drones are consumable, commercial-off-the-shelf products. An Amazon warehouse that has 20% of drones for hobbyists, but some are going towards military use, is “war-sustaining”, but no doubt contributes to the livelihood of all the workers—especially if you include the clothing, food, and other items stored in this warehouse. Again, I think attacking military supply chains is something most militaries would consider acceptable and legal, but this is an example of a supply chain that serves both populations.

• What about mobile networks and CCTV or Ring cameras used for navigation and target selection? A sensor used for military targeting is undoubtedly an object that makes a direct and effective contribution to military action. If those ring cameras are on houses in a subdivision next to a military objective, and the civilian deaths are proportional to the military objective, then the attack is legal. The International Criminal Tribunal for the former Yugoslavia (ICTY) when adjudicating the NATO bombing of a radio tower during the Yugoslavia campaign said as much, (75-76. Full decision here).

• What about Telegram servers? The same ones recruiting saboteurs, causing physical, kinetic damage that kills civilians? Returning to the ICTY decision, it may be legal “Insofar as the attack actually was aimed at disrupting the communications network.” But “disrupting government propaganda [and] such grounds alone may not meet the "effective contribution to military action" and "definite military advantage." This platform is both a civilian and military objective. If Ukraine did not have a vested interest in keeping Telegram up and running for its own purposes, I think the destruction of these servers would be legally debatable.

• What about a major economic lifeline of the enemy? Ukraine has consistently attacked Russian oil refineries. I am not sure this is legal under Hathaway’s solutions. Attacking civilian economic objectives is controversial in international law, and many nations won’t do it (She cites an example about U.S. allies not wanting to attack poppy fields in Afghanistan). While I think kinetic attacks on economic infrastructure should be put through her analysis in recommendations 1 & 3, I think if it is making enough of a contribution to be worthy of a sanction, it is probably worth enough to be bombed.

Protecting civilians requires consequences for the aggressor

These cases illustrate a tension between law and war and how these tensions are only going to increase as societies return to the norm of “total war.” I wish Hathaway’s piece were a promising solution, but I feel it is more of a prediction of things to come. It is hard for me to imagine an authoritarian military bent on subjecting a people to conduct this dual-use analysis. Even so, the reality is that for democracies to survive authoritarian incursion, there will have to be a full-scale civilian-military resistance, and this will make civilians a target, even under a strict reading of AP I Art. 52.

The solution is not to give up on international law, however. Democracies should continue a full-throated defense of the institution. They must refine these protocols and build on the initial research on dual-use objectives. I do not think a bright-line rule like “war-sustaining” analysis will work. Perhaps we should relook at proportionality for dual-use targets. Raising the threshold for acceptable collateral damage is one option. Another possible solution is to increase the burden of proof on the attacking army. Intelligence providing proof for attacking dual-use objectives should be retained, and a higher fidelity should be required.

Changing the law is slow. What doesn’t have to be slow are the consequences. All the laws in the world won’t mean a damn thing if war criminals and aggressive nations are not aggressively prosecuted at war’s end. And for there to be prosecution, there must be a war resolution where perpetrating leaders, officers, and government officials are turned over to war tribunals to be tried. And for this to happen, law-abiding nations must win the war. Without such a victory, there can be no justice for the civilians who have had the scourge of war thrust upon them.

Subscribe now

Share

Share Pioneering Oversight

In a move that flabbergasted a lot of policy wonks all over the country, President Trump OK’d the sale of Nvidia Chips to China, reversing a policy his administration implemented last April.

The administration said Nvidia would be allowed to sell the H20 chip after licenses are granted by the Commerce Department, according to the company. Nvidia said it would resume deliveries soon of the chip, which was designed for Chinese customers and has been a top seller in the country since 2024.

In addition, Huang said Nvidia has developed a new AI chip for China that he said would be useful for factory automation and logistics. (WSJ)

It seems at odds that the PRC, while being generally viewed as an adversary by many across the political spectrum, should receive a reprieve from the very commodity that may help them “win the AI race” with all the follow-on national security implications that come with that. at Consensus Drift has a good analysis of the possible reasons behind the policy change. His post is worth the full read, but here are his top-line points:

1. A Strategic Tradeoff? “The Trump administration may have judged that the risks of allowing China access to a downgraded AI chip were outweighed by the urgency of restoring U.S. access to rare earth magnets.”

2. AI Strategy Adrift? “By tying access to [export] controls to a commodity-for-chip swap, the administration risks reframing [export controls] as flexible instruments of diplomacy rather than principled tools of containment.”

3. A Return to Integration Over Isolation? “Rather than wall China off, many in the tech world have long argued that it’s smarter to keep China tethered to the U.S. tech stack. The idea is that even downgraded U.S. tools create dependence, and dependence creates leverage.”

4. A Concession to Industry? “…it may have been reframed internally as a necessary concession to stabilize a strategically critical U.S. firm (not its market cap but its competitive product position vis-a-via Chinese competitors”

Despite this move (unclear if there are winners or losers), the mistrust between the PRC and the U.S. persists. On August 1^st, “The Cyberspace Administration of China summoned Nvidia to address concerns that its H20 AI chips have backdoors, after a US proposal called for advanced semiconductors sold overseas to have tracking capabilities.” (Semafor)

Two other great articles on the Chip trade:

• The Rest of World’s piece on how Vietnam may come out as a winner in the U.S.-PRC Chip Trade War

• ’s AEI post on how the semiconductor trade works. This piece is also available on his Substack:

We don’t need no education

The conversations around AI in education are mostly centered around how to tell if students are cheating or if schools should be teaching kids “the skills of the future.”

What is rarely discussed, until now, is the complete outsourcing of teaching to AI. The problem includes inefficient teaching models, students being held back by peers, and the need for one-on-one bespoke tutoring. So, rationally, parents in Austin, Texas have pushed for an increase in school funding to hire more teachers and raise their salaries, thereby attracting more talent, and have launched a disruptive tech company outsourcing most of the teaching to AI that allocates 2 hours a day toward learning. (NYTimes)

This two hour learning model frees up students for other collaboration and “It is not a “screen school,” the founder insists.

This Austin school is hardly alone, however. The idea that more tailored help can achieve better results for students is not new, but others are saying that using chatbots comes at a cost—connectedness. This idea isn’t simply a tech-doomer mantra, but is being proven in psychological studies. When students exchange one-on-one coaching for chatbots, they lose social capital, which is essential for networking into jobs later in life, obtaining letters of recommendation, and forming social bonds. (The Markup)

China is also dealing with the same general education problems as the U.S.

Some schools, mainly urban ones, are over-resourced, while rural schools remain under-resourced. The PRC is hoping AI can level this playing field. Because the PRC has an exam-focused education system (progress to the next level is focused almost entirely on test scores, and teachers are evaluated on these as well, so naturally, teachers focus lessons to obtain the highest test scores), the AI experiment is leading to different results. Read this China Talk article for a deeper understanding of what is happening with AI and education in the PRC.

African airtime

Crypto is new, but trading in alternatives is not. Prisoners have used cigarettes as currency, a 6-pack and pizza will buy a friend’s help moving a couch, and I’m sure most millennials have used Pokémon cards to barter at one point or another. In Africa, “airtime,” a form of prepaid calling time, is a common currency. It began in the 1990s as a solution to sell cellphone services to the low-income continent, but continues to this day:

Today, more than two decades later, airtime is huge in Africa.

• In the first nine months of 2023, over ₦2.6 trillion (~$6.3 billion) was spent on airtime and data

• In Kenya, people spend KSh 3.4 billion a month on airtime.

• And the average South African spends R132 ($7.4) every time they top up their airtime.

[…]

According to the GSMA, over 70% of Africans have used airtime for something other than calls and data in the last five years.

• In Cameroon and the DRC, some churches collect tithes in the form of airtime.

• In Somalia, people send airtime to relatives in rural areas, who then sell it for cash.

• And some African businesses use airtime as rewards for surveys, referrals, and promotional programs.

This Tech Safari story is worth reading as it explores how airtime has become digitized and traded, and how other micro-payments work on the continent. If you enjoy this story or some other African technology stories I’ve written about, most of them are thanks to this newsletter. You can click on this referral link and sign up for Tech Safari directly.

In other news

• NIST Makes First Detection of Cannabis in Breath From Edibles

• Jury Says Tesla Was Partly to Blame for Fatal Crash. “The jury found that Tesla bore 33 percent responsibility for the crash, and blamed the driver, George Brian McGee, for the remainder.” Tesla could end up having to pay $243 million in damages.

• Citizen Scientists Are Accelerating Ecology Research, Study Suggests. “Thousands of scientific papers have used data collected by users of the platform iNaturalist, according to new research.”

• Henry Farrell posted a “Political Economy of AI” syllabus on his Substack. I’ve not gone through it all, but I’ve leaned on his work before, and he has a knack for distilling complex topics into manageable lessons. I didn’t see many non-Western sources on there, so in addition to Tech Safari linked above, I frequent “Rest of World” for a more global perspective.

From Firefighting to Future-Proofing

The email arrived at 6:47 a.m., and Ava was prepping her first cup of coffee of the day; she was going to need it. It was from the head of European operations, and it was not good. The Q2 compliance audits had revealed rising rates of regulatory citations, despite having just updated and re-launched the annual compliance training. Ava was the senior Training and Development Manager at TechFusion, a global technology company, and the increase in citations was very much her problem. TechFusion’s international support staff was vulnerable to increasingly costly mistakes and even disciplinary action from several significant government agencies. Ava’s role had always been equal parts strategist and firefighter. With over 5,000 employees spanning four countries (with further expansions on the horizon), her team was tasked with designing scalable, compliant, and effective training for a range of topics, including anti-harassment protocols, state-specific disclosure requirements, and enterprise software rollouts.

She had been feeling for a while that the traditional instructional design tools and mechanisms weren't quite keeping pace with the company’s growing size and scope. The audit results were the last straw; she needed to do something about this. Of course, the symptoms were clear; the cure, less so.

Ava learned three things that morning: 1) this was not a problem that could be solved by creating more slide-decks for the employee training center, 2) she was finally ready to seriously lean into building a comprehensive AI strategy at this company, and 3) this was going to be a long day!

The AI Playbook: Mapping the AI Journey

With her mind set on knocking this project out of the park, Ava devised a four-phase plan to completely overhaul TechFusion’s learning & development (L&D) program using AI. While committing to developing a companion compliance plan embedded in every phase, Ava’s project outline followed a fairly standard approach:

1. Root Cause Analysis and Organizational Needs

2. Research and Solutions

3. Project Planning and Implementation

4. Monitoring and Evaluation

However, integrating complex risk management, AI governance, and regulatory considerations into the project, particularly in an AI prompt, looked easier said than done.

Phase 1. Root Cause Analysis and Organizational Needs

Ava began with a thorough current-state workflow audit. She mapped out role-specific performance expectations vs. actual outcomes, cataloged all tools and resources in use, reviewed existing training curricula, and analyzed business metrics.

Some findings were not surprising: training modules lagged behind policy updates, translation bottlenecks slowed rollouts, and training fatigue was rising. Some of the data indicated a drop in certain high-value areas of employee performance, such as time-to-productivity for new hires and post-training performance ratings. However, crucially, compliance errors hadn’t improved, and performance metrics in this area of the business were trending downward, despite the company having required related training over the last few years.

Ava synthesized the results into a comprehensive needs analysis report. It highlighted three top-priority issues undermining TechFusion’s training efficacy:

• Generic, one-size-fits-all content that lacked role-specific or region-specific relevance.

• Slow development cycles for new training (especially when updates or translations were needed).

• Weak data analytics on training effectiveness, retention, and real-world behavior change.

In short, Ava realized she needed solutions to personalize learning at scale, accelerate content creation, and improve tracking of outcomes. Equally, she recognized that any such solutions must integrate with regulatory requirements and company values from the ground up.

Compliance Plan: In parallel, Ava conducted a high-level risk and regulatory scan to establish the compliance landscape from day one. She documented which global laws and policies governed employee training data, accessibility, and AI ethics at TechFusion. This included privacy laws like the EU’s General Data Protection Regulation (GDPR) and California Consumer Privacy Act (CCPA); accessibility requirements under the Americans with Disabilities Act (ADA) and Section 508 of the Rehabilitation Act of 1973; and keeping an eye on upcoming AI-specific regulations (such as the EU AI Act).

She also inventoried internal policies (e.g., data handling rules, code of conduct, HR protocols) that the new AI tools must adhere to. All these requirements were logged in a risk register, which Ava used to track ongoing issues and mitigation plans framed using industry best practices such as NIST’s AI Risk Management Framework. By flagging gaps and legal considerations now, Ava set compliance as a core success criterion from the outset, ensuring her AI project would be secure and lawful by design.

By the end of Phase 1, Ava felt confident that she had identified both the operational pain points and the compliance guardrails for her project, ensuring that any AI she introduces will be safe, fair, and legally defensible. In her mind, she wasn’t just planning a training revamp; she was laying a compliance-first foundation, turning potential legal landmines into guideposts for a smarter L&D strategy.

Phase 2. Research and Solutions

Armed with well-defined needs, Ava proceeded to research AI-based solutions. She deliberately mapped each identified need to a specific AI solution. Three promising ideas emerged:

1. An adaptive learning platform to dynamically tailor content to each employee’s role, location, and proficiency (addressing the generic content problem);

2. A suite of generative AI tools to help instructional designers speed up content development, auto-generate quiz questions, and instantly produce translations (addressing slow development cycles); and

3. AI-powered analytics dashboards to track learner engagement, quiz scores, and on-the-job performance indicators in real time (closing the analytics gap and allowing Ava’s team to identify and prioritize training needs quickly).

Ava sketched how these tools would fit together in a modernized “AI L&D stack” where each tool targeted a specific pain point, while being complementary pieces of one integrated strategy. She also factored in change management: planning early for user training and IT support so these solutions would actually be adopted by staff.

Compliance Plan: As she evaluated potential vendors and tools, Ava performed rigorous due diligence on each option’s compliance posture. She developed a checklist of questions based on industry standards, such as SOC 2 and ISO 27701. SOC 2, for instance, assures that a service provider has strong controls for security, availability, integrity, confidentiality, and privacy of data. ISO 27701 is a security framework that covers privacy management, essentially mapping to GDPR requirements for protecting personal data.

Ava required any AI vendor to supply independent audit reports or certifications demonstrating these controls. She also insisted on contractual protections: each vendor needed to sign a robust Data Processing Addendum (DPA) outlining how they handle TechFusion’s data in compliance with global privacy laws (Under GDPR, vendors are legally “processors” who must agree to protect data and assist the company in fulfilling obligations like individual rights requests.)

In addition, Ava sought vendors willing to offer indemnification against IP claims – an essential safeguard given the tendency of generative AI to reuse training data. For example, if an AI-generated training video accidentally infringed someone’s copyright, TechFusion wanted the vendor to bear the legal risk.

Crucially, Ava examined each AI solution’s approach to ethical AI, hallucination, and bias mitigation. She asked vendors for documentation of their bias testing history – for instance, had they evaluated their algorithms for demographic bias or accessibility barriers? One vendor demonstrated a robust internal policy: they retrained their models on diverse data and had a practice of human-auditing outputs for fairness. Another vendor, less prepared, had no good answer – a red flag in Ava’s book.

Additionally, each vendor had to support an audit trail for its AI decisions. For example, the adaptive learning system needed to log why it made certain content recommendations—a feature akin to providing a “right to explanation” for algorithmic outputs. Such transparency would enhance user trust and may be required under future regulations governing automated decisions.

By the end of Phase 2, Ava had systematically vetted and selected AI solutions that not only met her functional needs but also fit TechFusion’s risk appetite and compliance requirements. She chose vendors who could prove alignment with top standards (GDPR, SOC 2, ISO 27701, etc.), and who agreed to contract terms protecting TechFusion’s interests (IP indemnities, strong DPAs). This careful procurement process reflects a broader lesson: due diligence is crucial for successful AI adoption. As Ava noted from her industry research, organizations must evaluate third-party AI risks just as they do cybersecurity risks, by demanding evidence of controls and accountability. Ava’s thoroughness here meant her AI initiative started on solid footing, with Legal and IT stakeholders entirely on board. She was now ready to move fast in Phase 3, knowing that the AI tools she’d deploy were both cutting-edge and compliant.

Phase 3. Project Plan and Implementation

In Phase 3, Ava translated her strategy into a concrete action plan. She developed a 10-week implementation roadmap structured into five agile sprints, covering design, development, and deployment. Each sprint had technical objectives paired with explicit compliance goals, ensuring that building and integrating AI solutions would not outpace the company’s ability to manage risks. Her plan allowed for approximately two weeks per sprint and went as follows:

Sprint 1 – Data & Sandbox Setup (Weeks 1–2)

Prepare secure, isolated environments for developing and testing the new AI tools. Ava’s team “sandboxed” the adaptive learning platform and content generator – meaning they set them up on segregated cloud instances with dummy data. They also performed comprehensive data mapping for GDPR/CCPA, identifying what personal data (if any) the AI tools would use (e.g., job role, language preference, training history) and ensuring that data flows were documented.

Baseline security checks were completed, such as access controls, encryption in transit and at rest, and vulnerability scans on the new systems. By the end of Sprint 1, TechFusion’s IT team had verified that no real employee data would hit the AI systems until privacy safeguards and legal approvals were in place – a practice aligned with the principle of data minimization and NIST’s guidance to categorize and contain AI risks early.

Sprint 2 – Content Modernization (Weeks 3–4)

Begin creating new training content with the help of generative AI. Ava’s team took six outdated e-learning modules as pilots for a “refresh.” The goal was to achieve a 50% faster development cycle per module with AI assistance. However, no AI-generated material was published without rigorous human review. Ava instituted a dual-review process: each AI-generated slide deck or quiz question set was reviewed by a subject matter expert (SME) for accuracy and by a legal/compliance officer for appropriateness and adherence to policy. This dual SME+Legal review ensured that the AI didn’t introduce factual errors, biased language, or anything non-compliant (e.g., missing a disclosure or using non-inclusive terms). It echoes the emerging best practice that “Human-in-the-loop” systems are essential to maintaining safety and compliance.

Sprint 3 – Adaptive Pilot (Weeks 5–6)

Deploy the adaptive learning platform in a controlled pilot. Ava chose three groups (e.g., sales teams in three different regions) to trial the new personalized learning approach. Before scaling up, her team ran targeted bias and accessibility audits on the platform’s recommendations and content. They monitored whether the AI platform’s module suggestions were equitable – for instance, ensuring that the difficulty level and type of content recommended did not unintentionally vary by gender or other demographics in the pilot groups. Any signs of potential bias or “drift” were logged and investigated. Accessibility testing was also conducted with real users from the pilot groups who have disabilities, to confirm that the learning portal meets WCAG standards in practice. These audits reflect a proactive stance: rather than waiting for complaints, Ava wanted to identify and fix bias or accessibility issues early. The pilot itself was closely observed, and feedback (both quantitative metrics and qualitative comments from participants) was gathered as input for the next sprint.

Sprint 4 – Feedback & Fairness Hardening (Weeks 7–8)

Refine the AI tools based on pilot feedback and prepare for broad launch. One significant focus here was reducing AI “hallucinations” – the tendency of generative models to produce incorrect information. By tuning prompts and settings, Ava’s team aimed to keep any factual error rate from the content generator below 5%. (For context, state-of-the-art models like GPT-4 have a ~3% hallucination rate in controlled tests, so <5% was deemed achievable and safe for internal training content.) The team also verified that the audit logs and explanation features of each AI system were working. For example, they generated sample “decisions” from the adaptive engine (such as why a particular module was skipped for a learner) to see if the system could produce a meaningful explanation on request – a vital capability for transparency and user trust.

They ensured that all AI-driven actions were logged with timestamps and identifiers, allowing for the later tracing of any incident or anomaly in audit records. Essentially, Sprint 4 focused on fortifying the AI’s fairness, transparency, and safety before it was implemented company-wide. Ava also involved the AI vendors in this process, asking them to review the pilot results and assist in fine-tuning the models to address any issues uncovered (e.g., tweaking the content selection algorithm if one region’s users had lower quiz scores, to ensure it wasn’t a cultural bias in the content).

Sprint 5 – Enterprise Launch & Governance Handoff (Weeks 9–10)

Roll out the solutions to the entire organization and formally establish ongoing governance. In this final sprint, the adaptive learning platform was integrated with TechFusion’s company-wide Learning Management System (LMS) and made available to all employees. The new AI-generated training content went live, and managers were briefed on how to use the analytics dashboards to monitor their teams. Alongside the launch, Ava convened an AI Governance Committee – a cross-functional group including L&D, IT, Legal, HR, and executive sponsors – to take over steering the AI tools post-launch. The committee’s role is to meet regularly, review performance and compliance metrics, and make decisions on any necessary adjustments. Essentially, Ava “built in” a governance mechanism to sustain what she started. The committee also finalized an incident response playbook specific to the AI systems, so that if an issue arose (such as the AI recommending problematic content or a data breach occurred), there were clear protocols to pause the AI, inform stakeholders, and correct the course. Establishing such an oversight body aligns with expert recommendations on AI governance, as organizations should create internal committees or boards to ensure the ongoing responsible use of AI.

By the end of Sprint 5, Ava transitioned day-to-day oversight to this committee, though she remained closely involved as a member and as the L&D process owner.

Throughout Phase 3, Ava balanced speed with diligence. She kept the project on a tight timeline (a three-month turnaround for the pilot and launch), but thanks to embedding compliance tasks in every sprint, nothing critical was skipped, and each week had buffer time allocated for any “pause-and-patch” fixes if new risks were discovered – a lesson from agile software development adapted to AI risk management. This iterative approach resonates with the “Measure” and “Manage” functions of the NIST AI RMF, which emphasize testing, validating, and refining AI systems continuously to keep risks within acceptable bounds.^, Importantly, Phase 3 set up the structures (logs, committees, incident plans) that would make the AI solutions sustainable in the long run. Ava was essentially training the organization to carry forward the torch of responsible AI.

Subscribe now

Phase 4. Monitoring and Evaluation

With the new AI-powered L&D program fully launched, Ava shifted to Phase 4, focusing on continuous monitoring, evaluation, and improvement. She identified a set of key performance indicators (KPIs) to measure the project’s success on two fronts: business outcomes and compliance outcomes.

On the business side, she tracked metrics like employee course completion rates, average onboarding time for new hires, improvements in quiz pass rates, and reductions in help-desk tickets related to policy questions. Early results were promising – for instance, the average onboarding timeframe dropped from ~18 days to 12 days for employees using the adaptive learning path, and completion rates for mandatory training rose significantly.

Of course, Ava paid equal attention to compliance KPIs. Some of these included: Accessibility compliance rate, Bias parity score (an internal metric comparing quiz or performance outcomes across demographic groups to detect bias), security and privacy index (tracking encryption coverage, access log review frequency, etc.), and the number of incidents (any privacy breaches, AI malfunctions, or content issues reported). By formalizing these metrics, Ava ensured that the “definition of done” for her project wasn’t just adoption, but also ongoing adherence to ethical and legal standards.

Since TechFusion’s use of AI in training was new, they treated the first year as a pilot period requiring semi-annual updates to document any new risks or mitigations (modeled to be compliant with the Data Protection Impact Assessment {DPIA} requirements found in the EU GDPR). Furthermore, Ava’s team set up an automated bias scanning job to run monthly, using scripts to anonymize and analyze outcome data from the learning platform for any statistically significant differences between groups (e.g., comparing the distribution of quiz scores by region or gender). While this might have been beyond the letter of any law, it was in the spirit of emerging AI ethics guidelines and the “continuous monitoring” called for in NIST’s Manage function.

This vigilant monitoring quickly proved its worth. Within the first two quarters, two compliance incidents were detected and resolved through the new processes. In one case, the generative AI had inserted an outdated policy reference in a Japanese-language module – essentially a minor hallucination that slipped past initial review in Sprint 2. A Japanese sales manager noticed it and flagged it. Thanks to the audit trail, Ava’s team pinpointed the issue (the AI’s training data included an outdated policy document), corrected the content, and updated the prompt settings to prevent recurrence. The second incident was more sensitive: the bias scan alerted to a regional performance disparity – employees in one region were scoring on average 10% lower on a particular compliance quiz. The investigation revealed that some phrasing in the examples (written by the AI) was unfamiliar to learners in certain regions, which caused confusion. Again, the committee invoked the pause-and-patch approach: they temporarily withdrew the module, had regional experts revise a few questions, and re-released it after a fresh round of SME review. Because the governance framework was in place, these issues were identified early and addressed within a matter of days, with complete visibility provided by management. In regulatory terms, this demonstrated a strong “internal control” system and the ability to enforce the “accountability” principle of GDPR (which expects organizations to not only comply but be able to show their work in managing data responsibly).

By the end of Phase 4 (and about six months post-launch), Ava compiled an outcomes report for TechFusion’s executives. The results spoke volumes. The AI-driven L&D transformation had achieved significant efficiency gains and learning improvements while upholding compliance. Among the highlights: Training development time was reduced by over 50%, saving hundreds of work hours and yielding an ROI well above projections. On the compliance side, quarterly audits found zero major issues; all courses met accessibility standards; and no privacy breaches had occurred. In fact, one of Ava’s proudest achievements was noted in the report: an AI-generated micro-learning series on GDPR itself proved more engaging than the original human-made version, improving knowledge retention by 17%. Notably, it had undergone seven rounds of human and legal review before deployment. This underscored an uplifting message: with the proper checks and balances, AI can actually enhance compliance training (imagine that!). As TechFusion’s CEO reviewed Ava’s report, he was struck by how seamlessly the narrative of innovation was interwoven with the narrative of governance. Ava had not only delivered quantifiable business results but had done so in a way that reinforced trust – among employees, regulators, and leadership.

Appendix

This section contains two initial “deep research” reports used for this story. Reports have been fact-checked and iterated on before final publication in this post, but should be helpful for anyone exploring how to use the tool or wanting to see a potential end-product for their training programs.

Everything is downloadable on the Pioneering Oversight GitHub.

"AI Tools for Training Managers in Enterprise HR"

This ChatGPT "Deep Research" report was used as a starting point for this week's article. It identifies core use cases for AI implementation in the Learning and Development profession, typically a program-management role housed in a company's HR department. The Report explores compliance, ethics governance, and risk management considerations for AI-driven training solutions, as well as examining some possible future trends in this industry. For each section, ChatGPT was instructed to create short vignettes that highlighted the key points of that content. These vignettes were then used to develop the final narrative for the article.

"Comprehensive Risk Management & Compliance
Framework for AI in L&D"

This ChatGPT "Deep Research" report was created after drafting an initial outline for the story. Once there was a core storyline and "flow" to the article, this report was generated to fill out the narrative structure by providing a deep dive into the specific risk management and compliance aspects in the story. After the research portion, ChatGPT was instructed to rewrite a section from the original draft that specifically addressed risk and compliance. Although the final article underwent several more iterations, this rewritten section helped inspire how the risk and compliance content would be incorporated into the story.

Subscribe now

I am back in Poland after a two-week trip to Kyiv. It was my third trip to the Ukrainian capital and probably around my 10^th or so trip to Ukraine, although I honestly lost count a while ago.

In one sense, a lot has changed since my first trip in May 2022. The government is operating (from my perspective) pretty efficiently—not many legislatures can pass reforms to get into the EU and prosecute a major war against a former world power amid a revolution in military affairs. In another sense, it’s still the same country with the same resiliency that I encountered three years ago.

I wanted to put together some reflections, particularly around technology, innovation, and reform, but also about some of the broader global issues intersecting Ukraine, and about civil and cultural resilience.

The World is at War

The first couple of nights I sat in my Airbnb wondering if it was safe to dash outside to a shelter. I listened to Iranian Shahed drones and missiles assembled with Chinese parts, probably defended at some point by North Korean troops, descend onto Kyiv. Some of them I saw get shot down from my window.

Coupled with Ukrainian allied support across Europe and other parts of the world, it is hard to characterize this as anything other than a global conflict. Yet, rarely do you find people publicly characterizing it this way. It might not be a “World War” yet, but it is a “world at war.” The sooner allied nations understand this in a way that impacts their military spending and strategic decision-making, the more likely it will be to achieve a quick and decisive outcome.

Culture Resilience

The notion that this is a cultural war is often overlooked in war coverage that focuses on drone strikes and meters of territory gained or lost. Ukraine’s war is a defense of culture as much as a defense of territory. 20,000 Ukrainian children have been kidnapped and removed to Russia. That’s equivalent to about half of the public school population of San Francisco, Seattle, or Washington D.C. Telegram bots push Russian propaganda across the widely used app in Ukraine making their narrative literally inescapable.

The usurpation of Ukrainian culture is a story as old as the Tsar’s empire. From the Russians’ perspective, this isn’t about conquering a sovereign nation; it is about reclaiming a piece of lost territory for their empire. As one of my Ukrainian friends told me, “We’ve been fighting these motherfuckers for 300 years, why would we stop now?”

For an excellent history on the roots of Russian imperialism in Ukraine, I recommend two books by Serhii Plokhy:

• The Gates of Europe: A History of Ukraine

• The Russo-Ukrainian War: The Return of History

Daily life in Kyiv is part of this resistance. Statues have been boarded up and protected against Russian air attacks, and some valuable pieces of art and history have been removed from the country for safekeeping.

Despite the packaging up of monuments and statues to survive bombardments, Kyiv remains one of the most beautiful cities I have ever visited. I’ve not been anywhere in my travels where I could see a church from the 1100s, a 1929 Art Deco apartment building, and a Soviet brutalist government building, all in the same frame. Dilapidated buildings, either abandoned before the war or bombed out, are covered with a canvas façade. Posters, murals, outdoor events, and gathering areas are all over the city. It’s an amazing city to visit. Kyiv looks good.

These are important lessons in civil resilience. It’s easy for a cynic to walk around Kyiv and wonder why resources are spent on things like public spaces when the money could be spent directly on the war effort. It isn’t an either-or. The well-being and morale of the citizenry is the war effort.

In a democracy, people need to support a war, to know that they are fighting to protect something valuable and meaningful. The recent protests in Kyiv and then the government’s response are an excellent example of an active population in combat as well as civic life.1 In any society, morale and endurance matter. Just as you need to ensure your troops get adequate rest to be sharp for the fight, you also need to understand that the civilian population’s endurance levels require the same attention.

Legal Innovation and Reform

A department head on my last ship once suggested, “Whatever admin and meetings we would stop doing if we went to war, we should stop doing them now.” No one likes paperwork. I doubt combat makes it any more enjoyable. Yet, aside from having to learn and adapt strategically and militarily, Ukraine is also pushing through reforms in a way that should make the bureaucratically bloated EU and utterly dysfunctional U.S. legislature reflect. A couple highlights of their reform process:

• Ukraine is actively legislating to align with the EU, a requirement for joining. This process is onerous. If the EU changes its regulatory regime, Ukraine must follow suit, regardless of how recent their last reforms were.

• Despite this frustrating process, one EU representative told me as they outlined the problems facing Ukraine that Ukraine’s admission into the EU at this point is “irreversible.”

• Ukraine created special legal and tax frameworks for tech and defense businesses, allowing a more conducive infrastructure for companies to grow and scale production.

• A number of draft laws are circulating government agencies aimed at incentivizing private industry to work with the national government and protect entrepreneurs’ (including innovative soldiers) IP.

The Human Advantage

Here are some updates on Ukrainian IP developments and thoughts about how Ukraine’s expertise will spread after the war.

• Ukraine is seeking ways to compensate soldiers for their frontline innovations, either through royalties or lump sum payments. While the government would probably own the IP, this would be a truly novel approach to military innovation.

• Patents are problematic in Ukraine for several reasons, including government capture, inadequate legal enforcement, and tax regimes. This is part of the reason for the significant push for the reforms mentioned above, and it will likely be successful given the EU accession process and the increase in FDI interest.

• Patents are a problem globally because they take so long to acquire. It is not hyperbolic to say that 3 months of innovation in Ukraine is worth about a year of R&D in the United States. Why patent something when it is obsolete before the ink dries on the application?

• Trade secrets—which could protect innovation that was either unpatentable or that the inventor did not want to publicize—are also not conducive to the knowledge-sharing environment that is crucial to wartime innovation. There may be a financial incentive to hold your invention close to the chest, but there is a life-or-death incentive not to. This means a company’s financial well being (read, ability to scale and produce more drones) is stunted in ways companies in other sectors and other nations are not.

• A royalty-based or lump-sum payment, especially to soldiers, is probably good policy. It allows the government to acquire and distribute useful innovations and may keep soldiers happy with a little extra cash.

• Restrictive licensing is another regime that could be beneficial. The government could grant itself and Ukrainian-based companies a license to use the innovation inside Ukraine, but the soldier or entrepreneur has the exclusive rights to sell the innovation outside the borders.

The upshot of all of this is that what is essential in the Ukrainian innovation ecosystem isn’t so much the patents, but the human capital. Humans are innovating outside of a conducive IP protection regime, but also at a pace that makes traditional IP regimes obsolete.

When export controls are lifted, Ukraine is going to have an advantage over every drone manufacturer in the world because their gear is cheap and has been proven to kill the enemy. They will also have a massive informational advantage in building and using drones, and this knowledge should command a premium. If other countries want to catch up, they should plan to invest a significant amount of money to hire and retain Ukrainian talent.

Thoughts on news from the Rest of the World

‣Drone Dominance

My trip coincided with a significant amount of news related to Ukraine. The Hegseth memo on drones and a potential “Ukrainian-drones-for-US-Patriot-Missiles” were the big ones. I’m not going to rehash the Hegseth memo here, but I recommend those interested read this CSIS piece on what the U.S. can learn from Ukraine to “Unleash U.S. Military Drone Dominance” and my friend and business partner Jesse’s analysis on the memo.

Future // History

Signal // Noise 004

Good morning team - I’m writing from Brooklyn after wrapping up a month of intensive field work across Taiwan and Southeast Asia. The jet lag is brutal, but the clarity and insights into the key geopolitical and national security issues of our day is worth it…

Read more

9 months ago · 1 like · Jesse Nuese

‣Where Ukraine succeeds, the EU flounders

In trying to support startups across all sectors (which would undoubtedly help the defense sector the EU is trying to cultivate) the EU Commission had a plan to make a singular legal infrastructure for startups—allowing them to scale without having to navigate 27 different regulatory regimes. This unified regime is in jeopardy for various political reasons and its failure will be catastrophic for a unified European innovative defense-tech market.

‣Applying Differences to Lessons Learned from Ukraine

I have a friend (American) building drones in Ukraine, and one of the things we discussed is how decentralized yet effective the Ukrainian drone industrial base is. This was my experience as well when I toured drone factories: You go to a nondescript location outside a city that looks like an abandoned sawmill, and then, when you walk through the rusted sheet-metal doors, you find yourself in a highly sophisticated drone-making operation.

These are impressive operations, but it is essential to understand what distinguishes Ukraine from the U.S. and our allies when considering the next war, such as a potential conflict with China.

• The Ukraine method may be useful for forward-deployed U.S. troops, Australia, and East-Asian allies dealing with sporadic or potentially non-existent supply chains, but the ability for the U.S. to retool and centralize production if required should not be ignored.

• Decentralization needs to come with deconfliction. Up to 60% of Ukrainian drones are downed through "Electronic Warfare friendly fire." Deconflicting the EW space will be critical to militaries engaged in complex and coordinated, multi-branch, multi-domain operations.

‣Soft Power Fallout

One member from a European Union delegation told me that the fallout from the shuttering of USAID is not known and probably will not be known for a while. “Americans could move money and resources quickly to the small and medium-sized organizations that did some really good work.” No one can move capital as quickly or efficiently as Americans. This observation, unfortunately, aligns with a complete withdrawal of soft power from the Pacific region, where the U.S. has been described as “a disinterested and mercurial partner.” Having friends throughout the world matters—American soldiers and sailors may need a favor on a far-off Micronesian island one day.

Some Final Thoughts

At some point, the war will end. Just as politics changed on day one of the war, they will change again on day one of the peace. I wonder how much time the Ukrainian government is devoting to post-war planning. Compare World War II with Iraq: it matters, despite how premature it may seem.

As much as I wish there were a flood of FDI right now, I know the reality of the risk-averse private sector is that the spigot will turn on after a peace agreement is signed. I wonder if Ukraine will be able to regulate foreign investment in a way that is beneficial to Ukraine and supports a sustainable democracy.

I spent a lot of time thinking about what it will be like when the soldiers return from the east and the refugees return from the west. After one of the bombardments ended, I sat in my flat wondering if I could hear a Shahed or not—or perhaps it was just the AC or refrigerator. After 3+ years of drone attacks, people coming back from combat are going to have a whole new manifestation of PTSD the world has not seen.

I wonder if Russian sanctions will turn off and if the world will allow them to reset their economy as if nothing had happened. I doubt the world has the stomach for a Treaty of Versailles or even a Nuremberg, given that it doesn’t even have the stomach for increasing sanctions to a level that would reduce Russia’s wartime economic capacity.

What I have seen over the last three years is a nation and its people who are not just brave and resilient, but also adaptive and responsive: to their government, to their enemies, and to mercurial diplomatic relationships. I have faith that the Ukrainians will make the right choices and that their decisions will be their own, free, and self-determining.

Subscribe now

The PRC is leading the way in energy security by multiple metrics:

It’s pouring hundreds of billions of dollars into putting renewable sources like wind and solar on its grid, manufacturing millions of electric vehicles, and building out capacity for energy storage, nuclear power, and more. This investment has been transformational for the country’s economy and has contributed to establishing China as a major player in global politics.

Meanwhile, in the US, a massive new tax and spending bill just cut hundreds of billions in credits, grants, and loans for clean energy technologies. It’s a stark reversal from previous policies, and it could have massive effects at a time when it feels as if everyone is chasing China on energy.

It’s more than unfortunate that energy security policy has become a victim of the culture wars. The policy disasters go far beyond solar. Small modular reactors (SMR) could be a great renewable and secure energy source for the EU and the U.S..

[…] Small modular reactors (SMRs), which are essentially compact versions of nuclear power plants fit for industrial, urban, military, and other applications. They produce up to a third of the energy of a traditional, large reactor.

SMRs’ main advantages are that, due to their size, they can be transported either as whole units or in modules and later assembled at the final destination, thereby reducing the cost of installation from which conventional reactors suffer. Another advantage is that existing designs are based on passive systems to ensure their safety, thereby minimizing the chances of radioactive leaks.

Yet both the EU and the U.S. are not putting the money in the necessary places to secure this energy future. Fusion energy, another promising technological development, still lacks the required capital investments and falls far behind the PRC’s investment in the technology.

It is becoming increasingly apparent that, even if we were investing in energy security at the rate we should, the current administration’s immigration policy is driving away talent. Every time masked armed officers conduct a raid on a farm or Amazon warehouse, it scares hard-working people out of the workforce, and high-skilled labor takes notice.

American immigration policy, which used to attract the best and the brightest from around the world, is no longer the R&D haven it used to be.

Since the end of World War II, federal funding has helped U.S. companies dominate the cutting edge of computing, space exploration and medicine, delivering an economic tailwind for the nation. It made the country a dream destination for aspiring researchers, engineers and entrepreneurs from around the world.

Now that’s changing—quickly.

A March 2025 survey by the journal Nature of more than 1,600 scientists in the U.S. found that three-quarters have considered leaving the country. Respondents specifically cited the Trump administration’s hostility to scientific research and those who practice it.

Historically, three-quarters of international students who earn a Ph.D. in the U.S. have stayed long-term. America’s ability to retain these workers—who are not just highly trained but expensive to educate—has been one key to the country’s pre-eminence in innovation.

The America First Alone policy is doomed to failure. There is no reason that we should not be investing in talent and infrastructure domestically at the same rate that we import it. Making America inhospitable to the world’s best and brightest is to give away a strategic advantage unnecessarily.

Share Pioneering Oversight

Digital sovereignty, global self-determination

On Wednesday, France, Italy, Germany and the Netherlands signed founding papers for an organisation where they plan to collaborate on building sovereign digital infrastructure.

European public administrations are highly dependent on software and services made by US tech giants, such as Microsoft’s Office suite – with rising concern this makes Europe vulnerable to geopolitical shifts. (EURACTIV)

Digital sovereignty is the idea that a nation, or group of nations, controls the data, hardware, and software that they rely on every day. Think “friend-shoring” or “reshoring” for the internet. While this has generally been spurred by a difference in data privacy laws or regulations, the EU is clear that it no longer wants to rely on the U.S. or the PRC for its digital future.

This battle isn’t just in the EU, however. This Rest of World story highlights the African struggle for digital sovereignty. Here are some of the significant efforts, but the linked story has more on the obstacles.

• Most of the data gleaned from African internet users reportedly sits in data centers in Europe and the Middle East.

• In April, Nigeria asked Google, Microsoft, and Amazon to set concrete deadlines for opening data centers in the country.

• Other developing countries, including India, South Africa, and Vietnam, have also implemented similar rules demanding that companies store data locally.

• India’s central bank requires payment companies to host financial data within the country, while Vietnam mandates that foreign telecommunications, e-commerce, and online payments providers establish local offices and keep user data within its shores for at least 24 months.

• South Africa is the only African country where Amazon, Microsoft, and Google have built their own data centers[…] the South African data center market could attract investments of as much as $3.7 billion […]

Appeals court strikes down ‘click-to-cancel’ rule (The Verge)

A rule enacted by the previous FTC chair that would have given subscribers an easy way to cancel services was struck down in a federal appeals court, likely ending the rule altogether.

Last fall, industry groups representing companies that benefit from subscription revenue — including cable providers, entertainment studios, advertising companies, and home security firms — sought to block the rule in court, arguing the FTC aimed to “regulate consumer contracts for all companies in all industries and across all sectors of the economy.”

But in a unanimous ruling by a three-judge panel for the Eighth Circuit, the court found that the FTC under former Democratic Chair Lina Khan erred so gravely in its roll-out of the rule that it needs to be thrown out altogether. “While we certainly do not endorse the use of unfair and deceptive practices in negative option marketing, the procedural deficiencies of the Commission’s rulemaking process are fatal here,” the court wrote…The Eighth Circuit found that the agency skipped steps to issue the rule, including depriving stakeholders of “a notable opportunity to dissuade the FTC from adopting the Rule as proposed.”

Tech Policy Press published two worthwhile reads last week.

June 2025 Tech Litigation Roundup- This tracker includes not only a short synopsis of major tech cases, but also extended analysis at the end of the tracker.

How Free Speech Coalition v. Paxton Will Change Tech Policy- This is a good summary of the various political positions on either side of this case. It’s an easy read without a lot of the complex technicalities of legal analysis. While there are legitimate concerns about overreach on age identification laws—ones that can be sorted out in court—the lack of flexibility and policy solutions on the “anti-ID law” side is still troubling. I have yet to see a solution that opposes ID laws that protects children in the same way that age-verification does. I have also yet to see a convincing analysis that the harms, potential or otherwise, outweigh the negatives of the policy. If you have read anything like this that will change my mind, please send it my way!

Another free speech story this week:

France launches criminal investigation into Musk’s X over algorithm manipulation

Beware the Robots: AI-Enabled Sanctions Evasion is Here (Royal United Services Institute (RUSI))

This commentary provides an overview of AI-enabled sanction evasion. The two big takeaways:

1. AI is enabling adversaries to create shell companies rapidly, find loopholes, and alternative ways to comply with sanctions faster than regulators can keep up.

2. AI is allowing adversaries such as North Korea to bypass facial recognition technology and other hiring safeguards by creating synthetic personalities.

Some other cybersecurity news

Cybersecurity infrastructure, programs, and databases are all under fire in the U.S., making it difficult for the government to assist the private sector in combating vulnerabilities.

Subscribe now

Those 9% were doozies, no doubt, but fall mostly outside the scope of this newsletter. Here’s an update on some of the major cases and legal issues in the tech world:

Copyright cases

Anthropic and Meta both scored victories last week in their copyright infringement cases; however, these are just a couple of the numerous infringement cases still working their way through the courts.

In both cases, a group of authors…set out to prove that a technology company had violated their copyright by using their books to train large language models. And in both cases, the companies argued that this training process counted as fair use, a legal provision that permits the use of copyrighted works for certain purposes.

There the similarities end. Ruling in Anthropic’s favor, senior district judge William Alsup argued on June 23 that the firm’s use of the books was legal because what it did with them was transformative, meaning that it did not replace the original works but made something new from them. “The technology at issue was among the most transformative many of us will see in our lifetimes,” Alsup wrote in his judgment.

In Meta’s case, district judge Vince Chhabria made a different argument. He also sided with the technology company, but he focused his ruling instead on the issue of whether or not Meta had harmed the market for the authors’ work. Chhabria said that he thought Alsup had brushed aside the importance of market harm. “The key question […] is whether allowing people to engage in that sort of conduct would substantially diminish the market for the original,” he wrote on June 25.

It is hard to read too far into the tea leaves in these cases, but they are important as they will be an indicator of how to try future cases. Perhaps the Plaintiff needs to include additional types of proof or think of a novel approach to “market harm,” maybe something along the lines of unjust enrichment. Different jurisdictions with different laws can be tried, and of course, there is always an appeal.

Fuzzy suppositions

The U.S. legal system is increasingly resembling the legal systems of emerging markets I study. Legal interpretations and arguments are a realist’s utopia, and defendants often placate the powerholders rather than presenting their arguments on merit in a free and open judiciary.

Paramount officially bent the knee and kissed the ring by paying President Trump $16 million to settle a “defamation” suit:

Donald Trump just achieved a major legal victory over a media organization he sued for reporting he did not like. Paramount, owner of CBS, has agreed to settle Trump’s claim that “60 Minutes” violated Texas state deceptive trade practices law by selectively editing an interview with Kamala Harris to help her, and hurt him, politically. At stake for Paramount is a merger with Skydance media which, it reportedly fears, will founder on administration opposition if it does not settle the CBS claim. The importance of the deal to Paramount, together with the wish to avoid other retaliatory action by the administration, apparently outweighed all other considerations.

Trump now prevails on a more-than-dubious legal claim—characterized by one expert as “ridiculous junk” worthy only of being “mocked”—and demonstrates that he can use a combination of a personal lawsuit and implicit threats of federal government power to pressure a news organization into political submission. It also highlights the central role of civil society, within the media industry and elsewhere, in taking seriously—or not—an ethical responsibility for the defense of democratic norms and institutions.

Equally as frightening is the argument upholding the President’s continued refusal to enforce the TikTok ban, a law passed by Congress, signed by the President, and upheld by the Supreme Court UNANIMOUSLY. Evidently, because the law touches national security, the President has the prerogative to enforce it or not. The issue of a social media app may seem trivial, but when viewed from the perspective of “do we have a functioning democracy or not,” this is probably the most damning issue on the books.

International news

Australia, India, Japan, and the United States have launched a critical mineral initiative, coupled with tariff talks on technology. This may be especially helpful to India, as its promised lithium mine in Kashmir has proven to be a bust. India has also struggled to stay competitive in the AI race:

Despite its status as a global tech hub, the country lags far behind the likes of the US and China when it comes to homegrown AI. That gap has opened largely because India has chronically underinvested in R&D, institutions, and invention. Meanwhile, since no one native language is spoken by the majority of the population, training language models is far more complicated than it is elsewhere.

U.S. authorities have cracked down on a North Korean cyber criminal network funding their nuclear weapons program. “A group under sanctions and linked to North Korea allegedly stole about $620mn in a cryptocurrency hack in 2022, US prosecutors intend to show in an upcoming trial, illustrating its reach in digital currency.”

From Academia

Two new reports are out. I haven’t read them yet, but they come from two highly respected sources I frequently turn to for research:

A Primer on Russian Cognitive Warfare

Cognitive warfare is a form of warfare that focuses on influencing the opponent's reasoning, decisions, and ultimately, actions to secure strategic objectives without fighting or with less military effort than would otherwise be required. China, Russia, Iran, and North Korea increasingly use cognitive warfare against the United States in order to shape US decision-making.…[It] is much more than misinformation or disinformation.…Cognitive warfare is distinguished by its focus on achieving its aims by influencing the opponent’s perceptions of the world and decision-making rather than by the direct use of force.

Military and Security Dimensions of Quantum Technologies: A Primer

Quantum technologies are advancing rapidly from experimental research into strategic defence and security applications, fundamentally altering how information is sensed, shared and secured. Ground- and satellite-based quantum key distribution networks are already being deployed by China and the European Union, promising virtually unbreakable communication. Quantum sensing systems, capable of precise navigation without a global navigation satellite system as well as subterranean and underwater detection, are nearing operational use…This report therefore provides important background and recommendations aimed at supporting the creation of international ethical, legal and security frameworks that ensure quantum bolsters, rather than undermines, global stability.

In other news

Kids are making deepfakes of each other, and laws aren’t keeping up.

Laws are having a hard time adapting to the increase of minors making sexually explicit deepfakes of each other. Some states lack frameworks to address this type of behavior, while others impose zero-tolerance criminal penalties, even for minors. I tend to skew towards a more “actions have consequences” approach to crime, especially in this arena, but sending kids to jail for this behavior is akin to giving a 14-year-old a bottle of schnapps and keys to a car and then putting him in jail when he drives drunk. Instead of criminal responsibility, we should be preventing them from getting access to these tools in the first place—something that will require a strong, government interventionist approach, much like the prevention of serving alcohol to minors.

AI detection software used at universities is as ineffective as snake oil,

littered with false positives (particularly for international students), and easy ways to get around them. Chinese universities are screening essays with their version of the software, much in the same way American universities use the ubiquitous software Turnitin. The Markup wrote a phenomenal history on the software, highlighting a new concern: The software company is retaining all the essays professors submit to the software as training data. There have been a few successful pushbacks against this practice:

In 2007, high schoolers in Virginia and Arizona sued iParadigms, Turnitin’s parent company at the time, arguing that its database violated their copyright over their own writing. The courts disagreed. The students also lost on appeal, but some colleges still warn their faculty against using free online plagiarism checkers because of privacy concerns inherent in handing student work over to third-party companies.

The story also highlights a generational gap. One student interviewed said professors “tell students not to even use spell-check tools because they’re bolstered by AI. Microsoft Word, Google Docs and Grammarly now all rely on the same algorithms that create ChatGPT’s human-sounding responses to suggest improvements to users’ writing.”

As anyone who is in or has recently completed work at the university level can attest, the quote above exemplifies most academic dishonesty policies. AI in the education system isn’t a problem to solve, but a pedagogical one to adapt to. Instead of spending thousands of dollars on false deterrence, professors should revisit the goals of higher education in the first place and explore how they can achieve those goals with the aid of AI. If they are that concerned about cheating, then they can give in-person, handwritten exams or oral boards. This, however, would require a bit more effort on the part of the professor.

The new workforce is making a debut.

Digital workers (AI agents) have log-ins and bosses, just like human workers, at a few financial services and banks, and Amazon will have more robots than humans in its warehouse very soon. Since the ban on state AI regulation failed, upcoming state legislature sessions may have “human only” hiring laws to protect human jobs.

Share Pioneering Oversight

I’d love to discuss this project with anyone working in some of the areas or on some of the research questions I have listed below. Even if your work tangentially touches on these topics or falls outside my regional focus, I’d still love to hear about how you are thinking about these issues. Here are some initial questions I’m looking into:

• How do wartime controls and export regulations shape IP strategy for Ukrainian and non-Ukrainian companies?

• What barriers prevent Ukrainian IP from reaching global markets?

• What financing structures and mechanisms work for IP-rich companies?

• How are Ukrainian companies currently leveraging IP assets—licensing, sales, joint ventures, collateral?

• How do investors value and diligence distressed intangible assets?

• Can blockchain and decentralized networks enable secure IP transfer?

I’m also seeking expert perspectives for field interviews and collaboration in these practice areas.

• Ukrainian & International Law & Policy: Business, trade, technology transfer, IP, export controls, and regulation and compliance.

• Finance: Traditional/alternative financiers, IP commercialization experts, fintech.

• Industry: Tech and data-heavy industries, dual-use & mil-tech, medical, energy.

• Business: Founders/executives at any size company managing or creating IP.

• Regional focus: Ukraine, Central and Eastern Europe (including Poland, Estonia, and Romania), or any business-, tax-, or IP-friendly jurisdictions.

I’d appreciate it if you all could forward this information to your network. Please feel free to message me on this app, or reach out through email or LinkedIn. I think this work has great potential to help out innovative-rich but capital-strapped communities, not only in Ukraine but around the world as well.

Share

Pioneering Oversight’s Top-5 Stories From Last Week

1. Data Brokers Are a Killer's Best Friend

For-profit data brokers directly enabled the assassinations that occurred in Minnesota on June 14th. Multiple stalkings and killings have been enabled by these platforms, where, for under $100, addresses and other information can be purchased outright. This piece explains that there are some opt-out laws in various jurisdictions, but these do nothing to address everyday citizens who do not qualify for these protections. Besides, with so many data brokers out there, it would likely require a heavy-handed response preceded by political will to remove an individual’s location data from the internet for good.

2. Reddit in talks to embrace Sam Altman’s iris-scanning Orb to verify users.

“Reddit is considering using World ID, the verification system based on iris-scanning Orbs whose parent company was co-founded by OpenAI CEO Sam Altman.” This ID system uses iris scans to verify a user’s identity before creating an account or accessing a website. This article explains how the technology works, addressing many of the trust and security issues that plague other ID verification systems, as well as the understandable concerns about anonymity and privacy.

3. Trump extends TikTok ban deadline for a third time

The probably 180-word article linked in the headline is probably all you need to know: President Trump, again, refused to enforce a law passed by Congress, signed by the President, unanimously upheld by the Supreme Court. Some other good stories on the topic:

My piece on the background of the ban and the issues of free speech and national security:

Pioneering Oversight

Banning TikTok

This isn’t the first time I’ve written about free speech or TikTok, and given that the TikTok ban, barring SCOTUS intervention, is posed to take effect in less than a week, I imagine I’ll be writing on it quite a bit in the coming weeks…

Read more

a year ago · Kelly Crawford

This New York Times article on TikTok’s ties to the current administration

The president tried to ban TikTok in his first term but flipped his stance on the app last year — a shift that is credited in part to one of his donors, who has a sizable stake in ByteDance, as well as his own growing popularity on the app.

And this Executive Functions piece on the legal rationale (or lack of) for non-enforcement

Executive Functions

Trump’s Continuing Illegal Refusal to Enforce the TikTok Ban

Please click here to opt in to receive via email our Roundup—brief daily summaries of news developments and commentary related to executive power…

Read more

10 months ago · Jack Goldsmith

4. From Iran to China: Using Social Data to Detect Authoritarian Decline

This is an excellent and short piece on how to analyze resistance to authoritarian regimes. New technology can be a great help, and this piece lays out an argument and a brief history on how to approach this. Human intelligence, however, has always been and will always be a key component of the puzzle.

5. GENIUS Act

The GENIUS Act, a cryptocurrency regulatory bill, cleared the Senate this past week and is likely to clear the House as well. Two stories on the soon-to-be-regulated “stable coins” are worth a read:

1. The Rise of Stablecoins and How to Regulate Them. This piece provides a concise overview of cryptocurrency, stablecoins, their uses (both illicit and legitimate), and regulatory approaches in the U.S. and the EU.

2. This blog post by the law firm Paul Hastings gives an update on the key provisions of the bill, as well as updates on a host of different crypto-regulatory changes that happened this past week outside of the Senate.

   Share Pioneering Oversight

Subscribe now

The company filed for bankruptcy in March, and the bankruptcy court's decision to allow the auction of user data sparked outrage, congressional investigations, and lawsuits from 27 states and the District of Columbia. It is unclear whether these lawsuits are now moot, as the data will remain under Wojcicki’s control, albeit in a different entity. Nonprofit ownership may also alter the contours of the lawsuit. However, the case may turn on individual state laws. Virginia, one of the leading states in the lawsuit, for example, requires explicit consent for the transfer of genetic data.

A company’s control of user data after service has ended is problematic. For example, with social media platforms, users license their content to the social media company in perpetuity. Some “Terms & Conditions” (like Meta’s) have so many “exceptions” that in reality, they have broad discretion whether to keep the data or not after a user cancels service. Others (TikTok) explicitly say they have the right to your content and data forever. As companies rise and fall, merge, and restructure, this will likely be a hot area for future regulation and litigation. If anyone has expertise in this area—terms and conditions, perpetual contracts, or data management during Mergers and Acquisitions—I’d love to chat and collaborate!

More rules for rural broadband

BEAD was an ambitious Biden Administration program set to benefit millions of people, but it was a victim of self-imposed project-killing red tape. Four years later, as states have finally satisfied the rules enough to make a bid for the money, the rules change, creating a new, different set of hurdles for states to jump through. Last week, I posted a link to a John Stewart podcast that discusses the red tape snafu that has stymied the program for the last administration. Below is a post from the NCSL on the new Trump Administration rules, critiques, and how states are reacting.

New Rules for BEAD Change the Broadband Game

The June 9 release of new rules for the $42.5 billion Broadband Equity, Access and Deployment (BEAD) program intended to bridge the digital divide and bring broadband access to underserved Americans, particularly in rural areas, brought a range of reactions, from support to scorn. No longer is fiber the preferred delivery method, in favor of a tech-neutral model which opens the door to satellite-delivered services. Critics say the satellite models deliver slower service than fiber but others point to its affordability and wider access.

• NCSL State Legislatures News: States Are Banking on BEAD Funds to Fill Broadband Gaps

• NCSL: Broadband Legislation Database

• NCSL: Broadband Policy Toolkit

• Stateline: Trump’s broadband program overhaul favors Musk tech, strips low-cost plans

• Route Fifty: Feds unveil ‘critical reforms’ to BEAD program

Uncontrollable ads

Meta is suing Hong Kong-based Joy Timeline HK Ltd., “that ran ads on its services to promote an app that lets people create nonconsensual, sexualized images of others using AI technology.” Meta filed the lawsuit in Hong Kong, highlighting the globalized and sprawling nature of the ad market. Ad purchases on online platforms are automated and extremely fast, often subject only to automated oversight or user-generated flags, after the ad is aired. It may be technically infeasible to shut out a single buyer, or it may be that Meta just doesn’t want to. It is a unique feature of the ad-buying ecosystem that the only way Meta can prevent a customer from buying ad space on its platform is through an international lawsuit, rather than simply refusing them service outright.

Apple and Google continue to host PRC-linked VPNs on their app store.

“TTP’s April 1 report found that more than 20 of the top 100 free VPNs in the U.S. Apple App Store in 2024 showed evidence of Chinese ownership. None of these apps clearly disclosed their Chinese ties, and some obscured their origins behind layers of shell companies. Several of the apps were linked to Qihoo 360, a Chinese cybersecurity firm that has been sanctioned by the U.S. over its ties to China’s People’s Liberation Army, TTP found.”

Why does this matter? Well, PRC companies are required by law to share their data with the state, as well as being subject to all sorts of other intrusive CCP protocols. If a user intentionally routes their data to servers in China, then that data falls under PRC jurisdiction—whether they know they are intentionally routing it there or not.

As trade wars and export controls continue between the U.S. and the PRC, Chinese engineers are resorting to analog methods to circumvent AI and semiconductor export restrictions by smuggling data out on hard drives in personal luggage.

• Chinese engineers transport hard drives with AI training data to Malaysia to bypass U.S. chip restrictions.

• Chinese firms are shifting AI development to Southeast Asia, exploiting data centers with advanced U.S. chips.

• The U.S. is considering measures to prevent Chinese access to AI chips via overseas locations such as Malaysia.

From around the country

• Last year, there were $42 billion in class action settlements, and the ubiquity of technology is making them more prevalent. This story highlights ways you can keep up with which ones you are eligible for.

• Walmart and Amazon are considering creating their own stablecoin, a move that would reduce transaction fees and increase the speed of closing payments, essentially bypassing most of the traditional banking infrastructure.

• “Disney and Universal sue AI image creator Midjourney, alleging copyright infringement”

• A class action lawsuit and an FTC right-to-repair lawsuit against John Deere are moving forward. This case may have a ripple effect far beyond the farm equipment market, as wins will bolster the right-to-repair movement and hinder “predatory updates” that force consumers to constantly buy new products.

• “Minnesota Passes Groundbreaking Mental Health Warning Label Bill for Social Media Platforms”

• The Center for Security and Emerging Technology published a new report on U.S. space technology. “This report explores the U.S. advanced space technologies industry and highlights challenges and opportunities the state of the industry presents for national security.”

Technology & protests

There is a lot to write about the protests, particularly the ones in Los Angeles, and the role technology has played. I’ll leave a few stories and quotes I found particularly useful this past week:

• The Chatbot Disinfo Inflaming the LA Protests. A Wired podcast discussing the chatbots spreading lies about the protest

• Fake Images and Conspiracy Theories Swirl Around L.A. Protests. A great piece on the posts and conspiracy theories recycled from the 2020 BLM protests:

  “X posted a Community Note pointing out that the photograph [of a pallet of bricks] had nothing to do with the protests, but it still was seen more than 800,000 times. It was also widely reposted, including by several seemingly inauthentic accounts in Chinese.”

• Los Angeles Protests Amplified by Influencers and Online Creators.

  “[Influencers] don’t have a boss, so they can say they’re independent [reporters], but they’re following the exact same incentive structures as the conservative establishment media,” Mr. Callaghan [a journalist] said in an interview. “Click bait, key frames that don’t reflect the reality of the situation, doing whatever they can to absorb the audience that otherwise would be watching the same programming on Fox News.”

The protests in LA are, at their core, about immigration. The city with one of the largest undocumented communities is using grassroots apps from activists to evade ICE raids. According to open-source reports, these apps appear to be effective; however, their security protocols are unclear. Stepping away from the domestic cat-and-mouse game, the national intelligence community is concerned that foreign actors can exploit the data, as it essentially provides a map of where law enforcement officers are deployed. I haven’t used these apps myself, but conceptually, it doesn’t seem too different from Waze or Google Maps, which tell you where speed traps are located.

Data transfer obviously goes the other way as well. Federal Medicaid data has been sent to DHS that contains individual recipients’ immigration status, a policy that is certain to scare an entire class of people from receiving medical care in the future.

I’ll wrap up and leave you with a clip from a presidential primary debate 45 years ago. George H. W. Bush, one of the smartest and most qualified men to have ever served in the office, hits the nail directly on the head. And, I truly believe, speaks for more Americans today than it may seem—in tone and in content.

Subscribe now

1. Restoring American Airspace Sovereignty establishes the Federal Task Force to Restore American Airspace Sovereignty (Task Force). The Task Force will “review relevant operational, technical, and regulatory frameworks and develop and propose solutions to UAS threats.” The threats listed in the EO include:

“criminals, terrorists, and hostile foreign actors” “drug cartels [using] UAS to smuggle fentanyl across our borders, deliver contraband into prisons, surveil law enforcement, and otherwise endanger the public. Mass gatherings are vulnerable to disruptions and threats by unauthorized UAS flights. Critical infrastructure, including military bases, is subject to frequent — and often unidentified — UAS incursions.”

2. Unleashing American Drone Dominance aims to accelerate UAS technology within the broader American infrastructure. It orders that the Secretary of Transportation propose rules for a policy that enables UAS testing beyond the line of sight, create a framework for deploying AI tools to assist in UAS waiver applications, as well as other programs to further integrate UAS capabilities into the U.S. ecosystem, including Made in America policies and supply chain weakness analysis.

• President Trump is also set to use the Defense Production Act to “slash legal requirements - including some congressional funding approvals - relating to a law aimed at lifting U.S. production of critical minerals and weapons.”

Watch dogs

• Citing censorship of “conservative ideas,” “The Federal Trade Commission is investigating whether roughly a dozen prominent advertising and advocacy groups violated antitrust law by coordinating boycotts among advertisers that did not want their brands to appear alongside hateful online content.” X is the driving force behind this investigation. It’s possible that this is not a conspiracy, but rather a platform failing in the marketplace of ideas—and the actual marketplace.

• The June 19 deadline for the TikTok ban looms, although it appears the PRC social media app will receive another reprieve. For a background on the ban, you can read this post:

• Reddit is suing Anthropic for unjust enrichment, claiming the AI company failed to get a license for its data.

• The Commerce Department is revamping its rules to allow satellite internet funding (read Starlink, although there are others). The ease in regulations will allow satellite internet companies to tap into $42.5 billion. It will be interesting to see if this administration can accelerate the movement of funds to infrastructure, after the Biden Administration, with the help of special interests, failed spectacularly to implement its own internet infrastructure projects.

• For a great read on the federal courts, a highly influential district court is reviewing this administration’s actions; check out this Lawfare Article.

International news

• Australians may soon join the EU in being able to download applications outside of the Apple App Store. This is the latest in a slew of bad antitrust news for Apple. For an in-depth (and not yet dated) analysis of the antitrust lawsuits, you can read this past post:

• TikTok has agreed to ban the skinnytok hashtag that led children to eating disorder content. The hashtag ban doesn’t affect content, however.

• The EU launched a platform that allows policymakers (and industry) to identify areas with an influx of goods. “With EU trading partners imposing an increasing number of restrictive measures on trade” the risk for redirecting of goods and “dumping” into the EU market are higher. This is a tracker to keep an eye on. If it is successful, it could drastically change arbitrage markets for regulators and market participants alike.

Lastly,

It’s easy to be reactionary these days about elite schools. They are a crucial part of our R&D infrastructure, yet something doesn’t feel right about their multi-billion-dollar endowments that are invested in private equity funds, while at the same time, taxpayers subsidize research funding through grants and student tuition through government-backed loans. Bloomberg published a story about how these universities are hedging their bets and shifting their funds to avoid the wrath of the current administration. One figure that stuck out to me:

Yale’s portfolio grew from $1 billion to $41 billion since 1985 and is taxed at a rate of 1.4%. A House bill would change that rate more in line with other capital gains tax rates.

Subscribe now

Like them or love them, regulations are a part of our lives. A small business owner doesn’t have time to parse through what NIST standards they need to use—thankfully, Chris does that for us.

This first installment explores how one busy professional transformed a time-consuming side hustle into a streamlined, AI-assisted workflow without losing the “human touch” that made his work meaningful and unique. Here’s the roadmap:

• The story—A case study following “Chris,” a working professional who optimized his newsletter workflow using AI.1

• The results—Shareable assets summarizing the process and lessons learned.

• DIY Templates—A customizable, downloadable DIY toolkit with tool examples, risk mitigation tips, and blank entries (including ChatGPT Prompts) for your system. Curated by Chris, so you don’t have to!

A Day In The Life Of... A Professional With A Side Gig

The Juggling Act: A familiar challenge

The alarm slices through Chris’s pre-dawn haze at 5:45 a.m., a rude reminder that last night’s “quick polish” of his emerging tech newsletter turned into another midnight marathon.

His newsletter has a growing base of a few hundred subscribers, primarily professionals from across a wide range of industries, including legal, healthcare, finance, marketing, and energy. Some are curious about trends in tech like AI, blockchain, and VR/AR. Other readers work directly in these fields and use newsletters like Chris’s to stay up-to-date with rapidly changing policies, products, and regulations. He curates news, summarizes breakthroughs, and connects the dots, giving his audience a weekly digest that saves them time while keeping them informed.

This has been a passion project for over a year and is something he thoroughly enjoys, both because it keeps him up-to-date with these industries and because it adds value to his growing community. However, Chris is becoming increasingly worried about being stretched too thin between this side project, work, and family obligations (not necessarily in that order, of course).

Between his 8-to-5 job as a compliance officer at a local credit union, chauffeuring his daughter to her after school activities, and all the other honey-do’s, don’ts, and duties, his beloved part-time side-gig has turned into a full-time pain in the ass.

Scanning news, researching, outlining, drafting, editing, creating visuals, formatting, and posting takes a minimum of 7-10 hours a week, fragmented into early morning and late-night sessions. The time crunch was not only inconvenient but also becoming increasingly unsustainable.

Recognizing the risk, Chris decided it was time to lean into his content philosophy and practice what he preaches: leveraging AI to streamline this project and give him some much-needed breathing room.

Routine mapping

The first step was to analyze the existing process. Chris sketched out his weekly workflow to see where he was spending his time and break down the process into more granular tasks—each week followed a familiar but grueling pattern.

Monday & Tuesday:

Content collection. He’d comb through his list of newsletters, blogs, and industry sources, highlighting important information and saving relevant tweets and posts. That alone could eat up close to 2 hours.

Midweek:

The planning and curation phase. He’d review the saved list, group ideas by theme, and decide which stories were worth pursuing. He’d manually build an outline, pulling quotes, stats, and links into a running doc. This outlining and early research often took another 1.5 to 2 hours across several evenings.

Friday & Saturday:

Drafting. One long, uninterrupted push to turn that raw material into an actual, readable narrative. This was the most cognitively demanding step: structuring the piece, writing headlines, maintaining a consistent tone, and incorporating quotes and citations to create a compelling finished product. It often took 3 to 5 hours.

Sunday:

Formatting and visuals. Sunday was reserved for last-minute edits, creating the header image, checking citations, uploading everything to Substack, previewing the layout, formatting sections, and finally scheduling the post, as well as writing promotional blurbs for LinkedIn and Twitter. This final stretch could run another 1–2 hours.

All in, the process was slow, too spread out, and overly dependent on narrow time windows: early mornings before work or late evenings after bedtime routines. If anything slipped or took longer than expected, the entire process would back up.

Now that he had visibility into the full scope of his tasks and dependencies, he could begin looking for smart, targeted opportunities to streamline his processes. He was not trying to cut corners, but instead wanted to find ways to offload the repetitive grunt work that drained his focus and stole his weekends.

Tooling

AI tools are out there, but which ones does Chris need? The first step was to list every individual task in his weekly process, from curating sources, outlining, editing, and publishing final drafts. Next to each task, he added columns to track:

• Time required

• How cognitively demanding it was

• Dependencies or blockers

He then researched potential AI tools applicable to each task and captured the following information:

• Tool name and purpose

• Pricing tier (free, freemium, or paid)

• Learning curve or complexity to set up

• Whether the tool could fully or partially handle the task

Tasks like content gathering, summarization, and organization were ripe for automation. Others, such as tone management or outlining, could be assisted, but still require a human touch. Fact-checking and idea generation? No way he’d hand those off entirely.

Given his limited budget, Chris opted to prioritize free tools wherever possible and double down on what he already had, namely, his $20/month ChatGPT Plus subscription. Rather than expanding his tech stack, he committed to maximizing the value of this platform for outlining suggestions, summarizing sources, and drafting segments for review, while editing in-line as needed. Also in the mix were other general AI tools with free tiers, such as Google Gemini and Claud, as well as more targeted, yet affordable tools like Feedly for sourcing articles and content curation, Canva for visual and graphic creation, and Grammarly for copyediting.

With a mapped-out framework of his tasks, priorities, and available time, Chris carefully assigned relevant tools to each task, focusing on those that offered the highest value with the lowest friction. Because time was his scarcest resource, ease of use and immediate utility outweighed the depth of features.

Unfortunately, Chris knew that with great utility comes great responsibility. To comprehensively and responsibly update his system, he would need to perform a risk assessment and incorporate mitigation considerations into his overall plan.

Risks & compliance

Through his research, Chris found, unsurprisingly, a patchwork of resources, industry opinions, and proposed best practices. However, he finally settled on two robust resources that seemed to fit the bill perfectly: the NIST AI Risk Management Framework and MIT’s Risk Management Repository. Each had their unique super-powers and drawbacks.

He began to understand that the most significant and most immediate threats weren’t those of potential litigation and/or regulatory action, but rather the erosion of his audience’s trust.

If that trust eroded, everything else would follow. That’s why, even as he automated parts of his production, he kept a careful eye on the human side of the equation. He also made a note to consistently check in with these resources to maintain a robust governance and compliance program, especially if his newsletter started gaining traction, audience, and/or revenue opportunities.

The first significant risk he identified was loss of authenticity. If readers started to feel like the newsletter was machine-generated filler, they’d unsubscribe. So Chris doubled down on making sure his voice still anchored the content. Every AI-assisted paragraph was carefully reviewed and touched up, not just for accuracy, but for tone, relevance, and nuance. He also decided to add a disclosure to his publication, indicating when and how AI had been used.

Second came the risk of factual inaccuracy, namely that AI tools hallucinate. That’s why every stat, quote, fact, and reference surfaced by an AI was required to include an active link to the original source, which was then reviewed and verified by Chris. Chris also maintained a manual citation list and experimented with plugging his drafts into other AI tools, such as Anthropic’s Claud and Google Gemini, to act as secondary fact-checkers and ensure clean attribution.

Third, he worried about copyright. Using AI-generated images or relying on datasets he didn’t understand felt risky. He mitigated this by sticking with licensed tools like Canva and DALL·E and committed to regularly reviewing their commercial use policies.

As noted above, Chris referenced the NIST AI Risk Management Framework as a helpful tool to future-proof his setup. He didn’t need enterprise-grade compliance just yet, but he appreciated the simple structure:

1. Map the risks

2. Measure their potential impact

3. Govern with clear boundaries

4. Manage them with practical workflows

For now, that meant clearly documenting his use of AI and staying informed about future changes in disclosure rules or copyright law.

Conclusion: A better system

Once Chris had all the individual pieces of his plan in place, the next challenge was implementing those tools within a workflow that fully leveraged their strengths. This wasn’t as simple as swapping old steps for new ones. He had to redesign how each phase of the newsletter was executed to work in harmony with AI, not just around it.

He realized, for example, that attempting to brute-force a lengthy article draft through an AI model would often result in generic or bloated output. Instead, he restructured the drafting process into modular blocks that could later be refined and stitched together.

For research and outlining, he reorganized his approach to supply ChatGPT with batches of curated source material, allowing him to build topic maps and summaries that actually made sense.

As he “played around” with these new processes, he realized he was spending a lot of time providing additional instructions to the generative AI tools to get their output to better align with his expectations. After a lot of trial and error, followed by some better-late-than-never research, he came to understand how to better structure his prompts efficiently for on-point outputs. Accordingly, he developed a catalog of prompt templates, specifically tailored for each of the modular steps he had identified in his new workflow process. This turned out to be one of the biggest increases in the efficiency and consistency of the overall process, while also reducing the frustration factor back to a manageable level.

After several weeks of trial and adjustment, Chris established a workflow that enabled him to reduce his total newsletter workload from up to 10 hours per week to approximately 2.5 hours, saving almost an entire workday's worth of time each week.

From a monetary perspective, total tool costs came to less than $100/month, which was well within his allowable budget, on top of the time saved to reinvest in spending time with his family as well as the added value of having more bandwidth to think creatively, test idea’s, and grow his audience.

Now, Chris begins each week with an automated, but thorough AI briefing of trends and headlines. He spends midweek assembling and refining the outline, and wraps with a focused Friday night development sprint. Sunday is just for final edits, visuals, and queuing posts. No more late-night rewrites. No more skipped soccer games.

What was once a frazzled passion project has become a sustainable, creative habit. He even used AI to turn it into a podcast:

Project snapshot

Below are shareable assets that summarize the transformation of Chris’s newsletter workflow into a lean, AI-assisted system that reduces time, protects voice, and maintains trust—all while utilizing accessible tools and risk-aware strategies.

The primary tool Chris used was ChatGPT. Starting with the ChatGPT 03 model to assist with preliminary research after “plugging in” our initial outline. The 03 model was also used to generate a more comprehensive prompt, which was used to generate an initial draft based on the initial research and an updated outline. Once the sections were drafted, we opened up the full project in “Canvas” mode, switched the model to 4o, and began fine-tuning each section line by line. This process was extremely iterative and involved both instructing ChatGPT to make edits on its own, as well as making specific edits to language and organization through traditional in-person editing, then instructing ChatGPT to make updates based on those in-person edits.

Subscribe now

Appendix

This section contains all the tools and resources Chris used, including downloadable templates and easy-to-copy and paste blocks so that you can incorporate them into your daily content creation workflow.

Everything is downloadable on the new Pioneering Oversight GitHub.

Newsletter workflow template

This is a hands-on tool to help you plan and systematize your own newsletter workflow. Whether you’re building a side project, launching a company update, or running a thought leadership series, this template helps you:

• Break down the workflow into repeatable tasks

• Identify and compare AI tools that suit your budget and skillset

• Plan for legal and ethical risks before they impact your brand

Download it here.

Risk Planning Template

Before launching or automating your workflow, consider the biggest risks that could affect your credibility, compliance, or technical stability. Use this template to develop mitigation strategies and tailor them to your specific audience or legal jurisdiction.

This template includes common risks, example concerns, and suggested mitigation strategies. Customize as needed to suit your specific tools, audience, and business goals.

Download it here.

ChatGPT Prompt Template

Use the following prompt structure template to guide ChatGPT in helping you build your own newsletter system step-by-step, based on the format and workflow outlined in Part 3 above. The final copy-paste prompt is optimized for use with GPT-4o via ChatGPT Plus. It works best when used in conjunction with the Canvas feature, allowing for outlining and editing directly within one workspace.

Prompt Structure for ChatGPT (With Explanation)

1. Set the role and context.
Start by telling ChatGPT who it is (its role) and what you need it to do. This helps the model understand how to respond.

“You are my AI workflow assistant. I am building a repeatable newsletter production system and need help designing it step-by-step.”

2. Clarify the goal.
Be explicit about what you want to walk away with.

“I want to develop a complete newsletter workflow that is efficient, sustainable, and supported by the right AI tools.”

3. Break the request into manageable steps.
List the steps clearly and in order so that ChatGPT can walk you through them one at a time.

“Walk me through each of these steps and wait for my input before moving on:

1. Identify my workflow stages

2. Break down each into tasks

3. Recommend AI tools per task with cost and complexity

4. Suggest risk mitigation strategies

5. Summarize the system as a checklist.”

4. Prioritize usability and iteration.
Ask ChatGPT to be concise and wait for your response before progressing.

“Only show one step at a time. Keep it practical. Don’t move forward until I say so.”

5. Include constraints or preferences.
State your preferences clearly—e.g., budget, tech comfort, tone.

“I prefer tools that cost under $100/month total and don’t require coding or complex setup.”

Copy + Paste Prompt for ChatGPT

You are my AI assistant for designing a repeatable newsletter production system. Help me design a personalized, repeatable newsletter production process using a structured framework.

Walk me through the following steps, one at a time:

1. Help me define my weekly newsletter workflow (high-level steps like idea generation, research, drafting, publishing).
2. For each step, help me list the tasks I perform or expect to perform.
3. Recommend 3–5 tools for each task, with pricing, ease of use, and a short description. Include links.
4. Help me select the tools based on my budget, tech comfort, and goals.
5. Help me identify potential risks (e.g. AI hallucination, copyright, privacy).
6. Recommend strategies to mitigate each risk.
7. Offer a final recap of my new system in bullet point format.

Only show one step at a time and wait for my input before moving forward. Prioritize usability and real-world efficiency over novelty.

For a post-mortem of the U.S.-UAE deal, read this:

ChinaTalk

"Just Count the Server Racks"...

Zvi Mowshowitz of the Don't Worry About the Vase substack did an excellent job writing up the key dynamics of the deal. I’ll be running excerpts from his post with some comments of mine interspersed…

Read more

a year ago · 31 likes · 5 comments · Jordan Schneider

China warns US chip curbs could derail trade talks

Chinese officials warned Monday that Washington’s recent move to tighten export controls on advanced Huawei chips could undermine US trade talks with Beijing.

The US Commerce Department said last week that firms anywhere using the Chinese AI chips could face criminal investigations — Beijing said the measure threatened global semiconductor supply chains…

Washington’s targeting of Huawei could offer Taiwan more opportunities to strike deals with American chipmakers, an analyst told Bloomberg: At the conference Monday, Nvidia announced plans to build an AI supercomputer on the island and deepen its partnerships with Taiwan-based chip firms Foxconn and TSMC.

However, building in Taiwan and the billions of dollars worth of recent deals don’t seem to be enough for Nvidia. Nvidia CEO says US chip export controls on China are a ‘failure.’ Others and I have written extensively defending sanctions, and I won’t rehash all of that here, however, if the export controls were such a failure, it seems odd that the PRC is fighting so hard against them.

South Korea has legislation in the works modeled after the GDPR.

The legislation is also meant to prevent big tech monopolization and “to restrict dominant platforms from prioritizing their own products over competitors’ (that is, self-preferencing) and bundling products and services together. It also requires platform companies to include most-favored nation clauses to ensure that the terms provided to other countries also apply in South Korea.”

The current legislation, known as the Monopoly Regulation and Fair Trade Act (MRFTA), has drawn the ire of several major U.S. tech companies, including Google, Apple, and Meta. These companies have criticized the act as discriminatory and urged the U.S. Chamber of Commerce, industry lobby groups, Congress, and the U.S. trade representative to push back against it.

The U.S. seems particularly well poised to address the MRFTA in the context of bilateral trade. U.S Trade Representative Jamieson Greer stated on April 9 that President Trump would not permit “the European Union, or Korea, or any other jurisdiction set the rules for digital trade.” Prior to becoming the trade representative, Greer also argued, “I would urge caution on the part of Korean officials to avoid another clash. There is no good path forward where Korean regulations treat U.S. companies more restrictively compared to domestic—and Chinese—companies operating in the digital space.”

The U.S.-South Korea trade talks, in addition to a June 3^rd election, make immediate passage of the MRFTA unlikely, but not off the table entirely, as it will likely be an important chip in negotiations.

“The European Commission is creating a new category of ‘small mid-cap’ companies that will be exempted from eight EU laws.”

The eight laws in question that will be changed through targeted amendments are:

1. General Data Protection Regulation;

2. Regulation on Protection against Dumped Imports;

3. Regulation on Protection against Subsidised Imports;

4. Markets in Financial Instruments Directive;

5. Prospectus Regulation;

6. Batteries Regulation;

7. Critical Entities Resilience Directive;

8. Fluorinated Greenhouse Gas Regulation.

Google moves to reassure EU cloud users amid concern over Trump threat

Google is beefing up its “sovereign cloud” options in the EU, as US tech companies move to reassure the continent’s users that their access to crucial technology will be safeguarded at a time of escalating trade tensions with Donald Trump.

The Silicon Valley giant provides cloud computing offerings in Europe that ensures sensitive information remains on local servers and adhere to EU laws on data privacy.

Google told the Financial Times on Wednesday it was broadening these so-called sovereign cloud options, including a new “data shield” that provides additional cyber security protections to European clients.

After White House Clash, South Africa Sets the Stage for Starlink Approval

Officials on Friday began a policy review process that could pave the way for Elon Musk to bring his satellite internet service into South Africa for the first time without having to sell shares to Black South Africans.

South African law requires foreign companies to provide partial ownership to historically disadvantaged groups, which were prohibited from participating in many business opportunities during apartheid.

The law is meant to redress the economic inequality that persists three decades after that brutal system of racial segregation came to an end. Mr. Musk, who was born in South Africa, has called the laws racist.

“Starlink is not allowed to operate in South Africa, because I’m not black,” he posted on X in March.

For the record, it appears that Starlink is currently not allowed to operate in South Africa, not because Elon Musk is not black, but because he does not want to sell stock to Black South Africans.

Child Safety

• Trump signs Take It Down Act, criminalizing deepfake and revenge porn

• Eating Disorder Content Is Infiltrating TikTok.

“SkinnyTok” is being investigated by the EU and is the latest example of empirically documented harm aimed at children by social media companies.

“SkinnyTok” is a genre of videos about weight loss, ranging from “harsh motivation” to food diaries showing what women eat to stay “lean.” It has been compared to “pro-anorexia” and “pro-eating-disorder” communities that have thrived for over a decade on platforms like Tumblr.

A 2021 survey of 273 women who used TikTok found that 64 percent had been “exposed to disordered eating content” on the For You page of the platform. TikTok has since added a feature to show eating-disorder recovery resources to users who search terms associated with them, including “SkinnyTok.” The resources do not appear when searching for “harsh motivation,”

Lauren Breithaupt, a Harvard Medical School clinical psychologist and cognitive neuroscientist who specializes in eating disorders, says “patients are reporting specific TikTok and Instagram trends that amplify their anxiety or trigger disordered behaviors.”

A handful of EU countries are pushing to codify EU-wide age verification access and limitations for children’s social media access. Technical limitations remain: kids are using pictures of President Trump to bypass age restrictions on video games.

Also:

• Apple’s Tim Cook called Texas Gov. Greg Abbott to personally lobby against online child safety bill.

AI Moratorium

Some good analysis on the proposed AI moratorium law—a nugget in the House’s version of the budget bill that would preempt all state laws on AI for 10 years.

• On AI Policy, Congress Shouldn’t Cut States Off at the Knees

• Do We Need a ‘NIST for the States’? And Other Questions to Ask Before Preempting Decades of State Law

• House Republicans advanced a 10-year moratorium on state AI rules. California would be especially hard hit

In other news

• DJI drones are everywhere. The U.S. may still ban them

The most immediate threat of a ban comes from a clause in the 2025 National Defense Authorization Act, a U.S. law that sets the annual military budget. Passed in late 2024, it requires DJI to clear a national security review or face an automatic ban by the end of 2025. No review has yet been scheduled, and the legislation does not specify which agency should carry it out.

• AI scam factories force trafficked workers to defraud global victims

• Young Indonesians applying for tech jobs via Facebook and Telegram are trafficked to scam farms.

• Scammers use deepfakes, voice clones, and other technologies to dupe victims around the world.

• Americans alone lost $12.5 billion last year, mostly to investment scams.

And, two investigative long reads:

• 3 Teens Almost Got Away With Murder. Then Police Found Their Google Searches. “An arson attack in Colorado had detectives stumped. The way they solved the case could put everyone at risk.”

• We Made Luigi Mangione’s 3D-Printed Gun—and Fired It

Yet the simplicity is only surface-level. This administration has redefined what “friends and allied nations” mean to America, and is re-shaping how critical AI know-how and semiconductors move around the globe. The dealmaking kicked off with the Trump Administration scrapping the previous administration’s “AI Diffusion Rule,” a rule meant to protect American interests, security, and allies:

The Department of Commerce announced plans to rescind the Biden administration’s AI Diffusion Rule on Tuesday, just days before the new rules were set to take effect. The January regulation would have created a three-tiered system for AI chip exports: unrestricted access for 18 close allies, limited shipments for most countries, and complete restrictions for nations like China and Russia.

The trip began with a U.S.-Saudi Arabia deal:

• Donald Trump lauds Saudi Arabia as he unveils AI and defence deals. “US president highlights enduring ‘bonds’ with the kingdom as White House hails pacts worth about $600bn.”

Which was followed by a whole slew of private-sector deals:

• Among the biggest deals, Nvidia said it will sell hundreds of thousands of AI chips in Saudi Arabia, with a first tranche of 18,000 of its newest "Blackwell" chips going to Humain, an AI startup just launched by Saudi Arabia's sovereign wealth fund.

• Chip designer Advanced Micro Devices also announced a deal with Humain, saying it has formed a $10 billion collaboration.

• Qualcomm Inc, which said it signed a memo of understanding to develop and build a data centre central processor (CPU).

• Saudi Arabian firm DataVolt will invest $20 billion in AI data centres and energy infrastructure in the United States.

• Alphabet's Google, DataVolt, Oracle Corp, Salesforce Inc, Advanced Micro Devices and Uber will invest $80 billion in cutting-edge transformative technologies in both countries, the White House said, without giving details.

The UAE received similar deals:

• OpenAI to join US-UAE plan to build vast data centre in Abu Dhabi. “Deal on third leg of Trump Gulf tour hailed as ‘major milestone’ on road to American ‘dominance’ in the field.”

But everybody everywhere is hedging their bets, a sign that “American dominance” of AI will be harder to achieve:

• UAE partners with Italian startup for AI supercomputer, AT THE SAME TIME, EU Commission opens door for ‘targeted changes’ to AI Act. “Brussels is reviewing a swathe of laws in a deregulation push,” understanding that its regulations hamper investment. Case in point, this past week: “Intel [spared] with EU regulators over $421.4 million antitrust fine this past week.”

• Taiwan’s “GlobalWafers Plans Additional $4 Billion Investment in U.S.” AT THE SAME TIME these investments in semiconductor plants and AI data centers are being met with more and more “NIMBYism” across the United States.

• The United States is expanding controls on Huawei chips, telling companies that using Huawei’s artificial intelligence chips “anywhere in the world” could incur criminal penalties because they violate US export controls.” The U.S. is also considering adding more PRC companies to the export control “black list” AT THE SAME TIME “Nvidia plans Shanghai research centre in new commitment to China.”

The long-term security risks and strategic value are in question. The ease with which critical technology can flow out of the country, AT THE SAME TIME the U.S. is proposing to control essentially every good, service, and data packet entering the U.S. solely through Executive Branch prerogative, is befuddling.

Jordan Schneider has a pretty succinct wrap-up on the U.S.-Saudi-UAE deal and the seemingly non-sequitorness of our new foreign affairs posture:

These deals hurt even more, given that we may no longer have the R&D infrastructure to recreate this amazing innovation.

For the last century, America has built an impressive public-private R&D ecosystem. This ChinaTalk podcast gives a historical overview of the ecosystem, what it means to our competitive advantage, and what is happening now that we are gutting it.

With seemingly loyal-less large, private corporations OK moving critical technology manufacturing to every nation other than our democratic allies, and an administration content on doing the same, the answer may lie in private sector companies and research institutes that are still true-believers in democratic alliances—after all it is democratic tax dollars and democratic governments that subsidized and protected this phenomenal R&D ecosystem. The world is privatizing, and the great thing about private capital is its immense flexibility. Authoritarian capture of capital and emerging technology does not have to be a fait accompli.

Copyright shifts

Part 3 of the U.S. Copyright Office’s “Copyright and AI” guidance came out last week. The pre-publication report is 113 pages (a good chunk of it is technical background). Here are some key parts that you can jump to that will probably have the most significant impact:

1. Section IV.A.2.c (Analysis of Transformativeness) - Pages 45-48: This provides a legal analysis of when AI training is “transformative.”

2. Section IV.D (Factor Four - Market Effect) - Pages 61-74: This section examines potential market harms, including evidence of a growing licensing market.

3. Section IV.E (Weighing the Factors) - Pages 74-75: This section shows how the Office balances competing considerations, offering useful guideposts without predetermining outcomes.

4. Section V.C (Analysis and Recommendations on Licensing) - Pages 103-106: This provides practical guidance on how licensing might work, acknowledging that solutions may vary by context.

5. Section VI (Conclusion) - Pages 107-108: The final assessment and policy recommendations aim to provide a balanced approach, trying to support both AI innovation and creative industries.

In other news

• The latest version (as of May 18) of the House’s budget bill has a section that would “preempt” state regulation of AI. It is unclear if it would “preempt”, “eliminate”, or “pause” state-level regulation, if the section would be allowed under Congressional budget rules, or if it is even legal (The federal government can’t generally “delete” state laws). For an excellent summary of robust AI legislation states have passed, read this.

• In what would be a clear violation of U.S. law, terrorists have been paying for and receiving blue check marks on X, including:

An advisor to al-Qaida. One of the founders of Hezbollah. The head of an Iraqi militia group known for attacks on U.S. troops. And a top official with the Houthi rebels who recently lashed out at the “criminal Trump.”

• “Right to try” laws are gaining steam across the country. These laws would allow people who are terminally ill to try treatments not approved by the FDA. Once the Montana Governor signs the right-to-try legislation, the first “experimental treatment hub” may be in Montana.