Get you up to speed: Trump‘s heated reaction to Anthropic highlights power dynamics in AI safety.

What we know so far

In the most clear and consequential policy move on AI safety yet, the Trump administration has announced it will blacklist a leading AI lab over its refusal to allow unfettered access to its technology for military purposes.

It is the president and his secretary of war, Pete Hegseth, going nuclear over Anthropic’s refusal to allow the Pentagon to use its AI for “any lawful purpose”.

Describing Anthropic as a woke, radical left company, the US president said on his Truth Social platform that “The Leftwing nut jobs at Anthropic have made a DISASTROUS MISTAKE trying to STRONG-ARM the Department of War”, adding that the company’s actions were putting American lives and national security in jeopardy.

Until now, however, Anthropic was doing more than any other AI lab to support the Pentagon.

Anthropic’s Claude AI is the only frontier model already being used extensively for sensitive military planning and operations.

It’s been widely reported that Claude AI was used as part of the Pentagon’s “Maven Smart System” to plan and execute the military operation to capture Venezuelan President Nicolas Maduro in January.

The origin of the dispute wasn’t about Anthropic’s commitment to the US military; instead, its insistence on “red lines” in relation to the use of AI technology.

Anthropic’s CEO Dario Amodei demanded assurances it wouldn’t be used for mass surveillance of civilians or lethal automated attacks without human oversight.

In a statement on Wednesday, Amodei said some uses of AI are “simply outside the bounds of what today’s technology can safely and reliably do”.

In a post on X, equally as seething as the president’s, secretary Hegseth announced that, as well as being blacklisted, Anthropic would also be designated a Supply-Chain Risk – a legal intervention previously reserved for foreign tech companies seen as a direct threat to US national security.

Given growing concerns about AI safety, it’s a move that has shocked AI safety campaigners, but also raises serious questions about the future viability of the Pentagon’s “AI-First” strategy.

Secretary Hegseth has given Anthropic six months to remove its AI from the Pentagon’s systems. But there are now questions about what he might replace it with.

For the first time in the short history of superintelligent AI, the row appears to have united the AI industry.

In a memo to staff on Thursday, Sam Altman, CEO of OpenAI, which has also been in talks with the Pentagon, announced he shares the same “red lines” as Anthropic.

Separately, more than 400 employees at Google and OpenAI have signed an open letter calling for their industry to stand together in opposing the Department of War’s position.

In a copy of the OpenAI memo, Altman tells staff: “Regardless of how we got here, this is no longer just an issue between Anthropic and the DoW; this is an issue for the whole industry and it is important to clarify our stance.”

The move by the Trump administration appears, therefore, to be as much about power as it is about AI safety.

The Pentagon has already said it wouldn’t use AI for mass surveillance of the US population, nor unsupervised autonomous weapons.

Its furious response to Anthropic seems more in response to a big tech attempting to dictate terms to the government, rather than what those terms actually are.

In taking on Silicon Valley, which, though AI investment largely accounts for much of the current US economic growth, the administration has just declared war on a powerful opponent.