The US defence secretary Pete Hegseth [pictured] has threatened to remove Anthropic from the DoD supply chain unless the company agrees to allow its AI models to kill Donald Trump’s enemies.
However, it is unlikely that Anthropic would give a rat’s arse about that threat, as the DoD has grown dependent on its product and contributes little to the company’s bottom line.
The ultimatum follows a dispute over Anthropic’s refusal to grant unrestricted access to its models for classified military use, including domestic surveillance and lethal operations without direct human control.
During a meeting in Washington on Tuesday, Hegseth warned Anthropic chief executive Dario Amodei that the Pentagon could invoke the Defence Production Act, which would enable the government to compel the use of the company’s technology for national defence.
A senior Pentagon official said Anthropic had until Friday evening to agree to the terms, adding that failure to do so would result in the company being designated a supply chain risk. Anthropic said it was continuing good-faith discussions to support national security while remaining aligned with its usage policies.
All this would be a greater threat to Anthropic, but would not be the end of the story. It would involve court fights and long debates in Congress over whether the DoD could do this. It would arrive with the irony that Hegseth is claiming that Anthropic’s tech is strategically indispensable while simultaneously labelling Anthropic a supply chain risk.
The dispute highlights growing tension between the White House and one of the US’s leading AI labs. Anthropic has pushed for stricter AI regulation, while Trump and his advisers favour a lighter regulatory approach.
Trump’s AI adviser, David Sacks, has previously criticised Anthropic for what he described as regulatory fearmongering.
The Pentagon has relied on Anthropic’s Claude model for classified work through a partnership with Palantir, but is now in talks with rivals including OpenAI, Google and xAI to replace it.
