"What are you going to do, stop me? I've got the guns," is a wild government argument for tech pundits to support

Over the last couple weeks, Anthropic and the Department of War have been engaged in a war of words over something very exciting: terms of use. Anthropic, as a private company, has created a product and reasonably established terms of use for it. However, the Department of War doesn't want to adhere to all of them. They went so far as to threaten labeling Anthropic as a "supply chain risk", a classification rarely used and historically reserved for foreign companies suspected of being coerced by their governments to harm Americans.

Ironically, one of the fears has been that unlike the US, other countries will have governments that make demands on these companies they couldn’t afford to reject. Well, we’ve got a new administration in town, and it does things a little differently. With a Congress that won’t act and no other branch willing to push back, the executive can do whatever it wants. So when the federal government wants to use a private company’s property for its own purposes, it does what those other countries do: compels that company with every bit of force it can muster.​​​​​​​​​​​​​​​​

That's how we got to last Friday, which was the deadline the Department of War set for Anthropic to cede control of their product in the way the federal government demanded. And Anthropic, to their credit, said no. This, of course, sent the federal government into a tizzy. They had a hissy fit, and now they're mad at Anthropic. Who knows what's going to happen, because you can't take anything they say seriously, but it's not great right now.

One thing I almost posted to social media on Friday night after I heard the Anthropic news was a snarky comment about how I was sure Ben Thompson would have an article up on Monday about how the federal government is totally right. I thought that was a little mean, criticizing someone for something they hadn't done yet, even though I suspected they might. So I didn't post it. But Monday morning, I wake up, there's a new issue of Stratechery in my inbox, and yep, it is an entire post about how it's actually good that the US Department of War is mandating that a private company do whatever the United States government says, and if they don't, they will make them.

There are 3 things that are notable about this to me. First, it's an on-its-face absurd turnaround for someone who has spent the last decade aggressively defending the rights of US companies against the federal government. When Biden administration staffers sent links to tweets to Twitter, suggesting they broke Twitter's terms of use, Thompson was outraged that the government would dare even suggest that a company follow the rules they had set for themselves. Now, the literal head of the Department of War and the President of the United States are telling a private company that they must change their product to work a certain way, and must update their terms of use to allow the US government to do anything legal, which, because there are no laws around AI, is basically everything (as Thompson points out near the end of his post). His position is that this is actually reckless from Anthropic, that they should update their terms to allow anything the government wants, and that the government is being consistent with American values. In short, if the government tells you how to build your product in a way that kills or surveils people, companies must comply. If the government tells you to moderate content a certain way, companies must resist.

It's worth pausing on that "anything legal" framing, because it's doing a lot of heavy lifting. Thompson (and freaking Palmer Luckey) appeals to democratic accountability as the reason Anthropic should defer to the government. But where is the democratic process here? Congress has passed zero laws governing the use of AI in military applications. There are no statutes defining what "fully autonomous weapons" means, no legislation setting boundaries on AI-driven surveillance. The Department of War is not invoking a law passed by elected representatives. They are issuing demands backed by executive power, full stop. Thompson frames this as the democratically elected government asserting its rightful authority, but that framing only works if you skip over the part where the elected legislature has done literally nothing on this topic. What we actually have is an executive branch acting unilaterally, which is precisely the kind of unchecked government power Thompson has historically warned against.

For what it's worth, it's hard to pin Thompson down on these things sometimes. He's railed against the "paternal" federal government making any choices for adults, and then commented on a recent Dithering episode that he thinks the government should be more "paternal" (his words) when it comes to gambling.

The second part of this is that, while the inconsistency is striking, it's not entirely surprising. Like I said, I already had the post ready to go on Friday predicting that this would be his take. Over the years, for all of the great business analysis Thompson does for tech, I've seen a pretty clear pattern in how he covers different administrations. During the Biden years, he took every opportunity to criticize that administration, naming Joe Biden by name and pushing back on basically every tech-related policy. Since the start of 2025 and the second Trump administration, the tone has shifted noticeably. As someone who listens to the Stratechery podcast every day, I have not heard him direct anything close to equivalent criticism at this administration. He has a tendency to steel man anything it does.

It's not just this case, either. When Trump issued seemingly random tariffs that weren't based on any sort of economic theory and were more random gestures to different countries who he had beef with. When those happened, he was one of the few voices out there saying these are actually not as bad as people think, it's actually 4D chess, it's actually really clever what he's doing. The second one I can think of is last year when the FCC was very clearly telling CBS that they would lose their license if they didn't take care of the Jimmy Kimmel situation, if they didn't limit Jimmy Kimmel's speech. And do you know what Ben Thompson's article was that week? It was something critical of the Biden Administration, of course. Like with all other speech squashing over the past year, Thompson has nothing to say about current events, he only has interest in relitigating the past.

I'm not trying to read Thompson's mind, and I could be wrong about what drives these editorial choices. We all have our biases, myself very much included, and I view this situation through mine. But Thompson presents himself and his platform as someone who gives you the analysis straight without a political thumb on the scale. The pattern I think is so clear to see makes that hard to take at face value.

Which brings us to the third thing I find notable. Thompson frames his entire argument around a concept he states plainly:

International law is ultimately a function of power; might makes right

And he closes the piece with:

I don't want that, and, more pertinently, the ones with guns aren't going to tolerate it. Anthropic needs to align itself with that reality.

"The ones with guns" here refers to the US federal government, by the way. Thompson would argue he's being descriptive, not prescriptive, just laying out how power actually works. But you can't spend an entire piece arguing that Anthropic should submit, present only two options (submit or be destroyed), and then claim you're merely observing. That's an argument dressed up as analysis.

And the argument itself has a name. "Might makes right" is the language of authoritarians, or as Edgar Temam wrote:

The first principle is the common meaning, namely, that the dominance of the mightier over the weaker is right. This principle is generally considered to be not a definition of justice but an expression of injustice.

It's also, plainly, the language of the schoolyard bully. "What are you gonna do, stop me?"

Thompson presents 2 options going forward for this situation:

Option 1 is that Anthropic accepts a subservient position relative to the U.S. government, and does not seek to retain ultimate decision-making power about how its models are used, instead leaving that to Congress and the President.

Or:

Option 2 is that the U.S. government either destroys Anthropic or removes Amodei.

So either the federal government mandates how Anthropic builds and sells its products, or the federal government destroys the company or removes their leadership and (presumably) inserts their own supporters in his place. The classic small government process…

What I think is really telling, and is what Thompson has critiqued others for in the past, is how his entire argument rests on an unproven prediction of the future.

Current AI models are obviously not yet so powerful that they rival the U.S. military; if that is the trajectory, however…

His entire premise is that AI is an even bigger national security risk than nuclear weapons, and therefore the U.S. government must have absolute control over them. It seems almost comical I need to say this, but this has not been established by reality yet. Large language models are undeniably an impactful piece of technology that will have some impact on the world going forward. But to suggest that they are more critical to national security than nuclear weapons is quite a stretch, and is regulating today based on what possibly maybe could be true in the future. This is exactly the sort of argument that Thompson routinely criticizes anti-monopoly people for expressing, but here he goes with it when it's convenient for him.

And this brings us full circle to the democratic accountability question. Thompson argues that Amodei is "unelected and not accountable to the public," and that decisions about military capabilities should be left to Congress and the President. Fine. But if that's the standard, then where is Congress? If AI really is as consequential as Thompson and Amodei both claim, then the absence of any legislation is the actual scandal here, not a company maintaining its terms of use. You can't simultaneously argue that democratic institutions should govern these decisions and then cheer when the executive branch bypasses those institutions entirely. If the answer is "Congress should pass laws," then the correct response to Anthropic's terms of use is to pass laws, not to threaten a private company with destruction.

As I said above, I think Ben Thompson provides valuable business insights. However, I think he's drank the Kool-Aid on LLMs, he presents his writing as immune from political bias even though the pattern suggests otherwise, and I think all of his worst tendencies came together for this one no good, very bad post.