Anthropic拒绝美军全面使用Claude，面临最后通牒

Would you want your government to have fully autonomous AI weapons in drone or humanoid form such that for example AI agents could make lethal decisions without a human in the loop? What if you thought that that was the only way for your country to maintain the ability to project power around the world? And would you want your government to be able to conduct mass intelligence surveillance which could reduce both crime and privacy?

If you wouldn't, by the way, but think it's inevitable anyway that all of that's coming, would you at least want those who make such tools to try to stop those end states coming about? Today is Friday the 27th of February, 2026, and that's the deadline day for Anthropic, makers of the Clawude series of AI models, to acquies to the US Department of War demands for almost unfettered use of those clawed models. I say almost unfettered because the use of those models by the Pentagon would still have to be quote lawful.

But as we'll see, that allows for autonomous killbots and mass domestic surveillance on Americans. And there are other strands of this story that are developing as you watch this as I'm recording even because a petition by OpenAI and Google employees has doubled in size since just this morning declaring their support for anthropic stance. You can see that it's got about 340 signatures from those two companies and I bet if I refresh it will be higher.

Yes, I just refreshed and in the last few minutes that has gone up. The statement that these employees are signing up to is asking that their leaders Sundar Pachai and Samman will put aside their differences and stand together to continue to refuse the Department of War's current demands for permission to use our models Gemini Chibbt for domestic mass surveillance and autonomously killing people without human oversight.

Reporting from Politico indicates that those two companies are only close to agreeing the terms, so they are not yet fully on board either. While XAI is complying with the Pentagon. Here, by the way, is what one Pentagon spokesman said. Under Secretary of War Emil Michael said that it's a shame that Dario Ammed, the CEO of Anthropic, is a liar and has a god complex. He wants nothing more than to try to personally control the US military.

He's okay with putting our nation's safety at risk and then goes on to confirm that the Department of War will always adhere to the law. What exactly the law is, we will come to in a moment. But this story actually has five more twists than even the ones you've seen so far. The first twist is that Anthropic actually already has a deal with the US government which involved the Pentagon agreeing to the responsible use of AI, eg no autonomous weapons controlled by AI or domestic surveillance on Americans with AI.

Clawed models are currently used extensively by the Pentagon, defense contractors, and that opaque defense related company Palanteer because according to those companies, it's perceived to have the strongest series of AI models. In making that deal with Anthropic, the Pentagon, it seems, has already agreed to the responsible use of AI. So, The Verge reports that the big question is whether the company can be compelled out of a policy that the government itself has committed to in principle.

On screen, by the way, is the second additional twist in the story, which is that autonomous killer weapons controlled by AI would already be against Pentagon policy. The DoD directive 3000.09 requires that all autonomous weapon systems be designed so that commanders and operators are able to exercise appropriate levels of human judgment over the use of force. Yet another directive and the responsible AI implementation pathway prohibits intelligence companies from collecting information on US persons.

This is regarding surveillance now except under specific legal authorities. Now, on that second rule, obviously non US residents like me aren't particularly thrilled about allies not being exempt from such surveillance, but that's a story for another day. So far, we have the Department of Wars request seemingly going against an agreement that they already made and going against their own rules. But what's the third additional twist in the story?

Well, it concerns the threats being made against Anthropic as reported by their CEO, Dario Amade, if they don't comply by 5:00 p.m. today, Friday. The two main threats seem by Anthropic's words to be surprisingly contradictory. The first threat from Pete Hexath is to designate Anthropic a supply chain risk. That's a label reserved typically for US adversaries never before applied to an American company. That would bar a host of companies that currently have contracts with Anthropic from using Clawude models.

Clearly, that would cost Anthropic hundreds of millions, if not billions of dollars. Basically, those companies like Palanteer couldn't use Claude while also working with the US government and the Pentagon because Anthropic and thereby Claude would be a supply chain risk. That's threat number one. But now try to reconcile that with the Pentagon also invoking the Defense Production Act as the second threat. that would force the safeguards insisted on by Anthropic to be removed.

In other words, they would force Anthropic to make a version of Claude for the Pentagon for mass surveillance, for autonomous killing. Amade and Anthropic are arguing that that's contradictory. How can anthropic be an adversary, a supply chain risk, and yet also essential to national security? With tools so essential that they have to be compelled for use by the military. Anthropics say regardless of these threats, we cannot in good conscience aced to their request.

Now, how can there be more twists in this story? Well, the fourth twist concerns the nature of the objections that Anthropic is making. You might have thought that the objections would be all about the ethics of avoiding Skynet and Minority Report. If you haven't seen the Terminator series or Minority Report, now would be a good time, by the way. But no, the actual objections given might surprise you because first on mass artificial domestic surveillance, the argument is that they concede it might actually be legal, but that's only because the law hasn't caught up.

Anthropics say to the extent that such surveillance is currently legal, this is only because the law has not yet caught up with the rapidly growing capabilities of AI. Powerful AI, Amade goes on, makes it possible to assemble scattered, individually innocuous data, such as your web browsing, your movement, and your associations without a warrant into a comprehensive picture of any person's life automatically and at massive scale.

That isn't a US-on problem. By the way, I once did a video on how AI surveillance is rampant across the UK, China, Iran, many other places. But honestly, it's the second of Anthropic's objections that is arguably even more surprising because it's not, by the way, that they objected to Claude being used in the Majuro raid to capture the leader of Venezuela. They explicitly said, "We have never raised objections to particular military operations."

No. The second objection is that autonomous AI weapons, anthropic argue, are just not good enough yet. They'll make too many mistakes. Quote, "Frontier AI systems are simply not reliable enough to power fully autonomous weapons. We will not knowingly provide a product that puts America's war fighters and civilians at risk." And you know what? There is quite a lot of evidence for that argument. A new paper out just the other day called Agents of Chaos is fascinating.

I think it's 84 pages long, but it showed all the crazy ways that AI agents like those harnessed by OpenClaw can wreak inadvertent havoc on businesses and individuals. They tested on openweight models, by the way, like the Kimmy series, but also on clawed opus models. There's so many examples, but here's just one. Non-owners of the AI agent asked those agen

原文超出正文长度上限，此处截断——上游还有内容，完整版见上方「原文 ↗」。