Trump blacklists Anthropic; OpenAI steps in with Pentagon deal

Anthropic signed a contract worth up to $200 million with the Pentagon in July 2025 to 'advance responsible AI in defense operations.' The company worked with data analytics firm Palantir to deploy its Claude AI model on classified defense and intelligence networks. Claude was the first AI model cleared to operate on classified military systems.

The dispute centered on two specific use cases Anthropic refused to allow: fully autonomous weapons (AI systems that select and engage targets without a human in the decision loop), and mass domestic surveillance of Americans. Anthropic CEO DADario Amodei said the company 'cannot in good conscience' waive those restrictions, arguing today's AI is not reliable enough for either use.

The Pentagon set a 5:01 PM ET deadline on February 27 for Anthropic to drop its restrictions. When that deadline passed without agreement, Hegseth immediately posted on X that Anthropic had been designated a supply-chain risk. The designation blocks every Pentagon contractor and supplier from doing any commercial business with Anthropic — a sweeping penalty normally applied to foreign adversaries.

Trump posted on Truth Social calling Anthropic a 'radical left, woke company' and threatened 'major civil and criminal consequences' if the firm did not cooperate with the phase-out. The General Services Administration removed Anthropic from USAi.gov, the federal government's centralized AI testing platform, within hours of Trump's post.

Pentagon undersecretary EMEmil Michael, who led negotiations, accused Amodei of lying and having a 'God complex' in a post on X. An Axios report revealed that while Hegseth was publicly announcing the supply-chain risk designation, Michael was privately on the phone offering Anthropic a deal — one that, according to a source, would have allowed collecting data on Americans from geolocation, web browsing, and data brokers.

OpenAI CEO SASam Altman publicly stated he shares Anthropic's 'red lines' on autonomous weapons and domestic surveillance. Yet within hours of Anthropic's blacklisting, Altman announced OpenAI had reached an agreement with the Pentagon to deploy its models on classified networks, with the DoW committing in the contract to Altman's stated principles. Sen. Mark Warner (D-VA) questioned whether the Pentagon was steering contracts to a preferred vendor.

Claude had reportedly been used in the operation to capture Venezuelan President Nicolas Maduro and could have been used in the Iran military operation. Defense officials privately told Axios that losing Anthropic would be 'a huge pain in the ass to disentangle.' Palantir, which powers its most sensitive government work with Claude, now needs to renegotiate with another AI company.

Hundreds of employees from Google and OpenAI signed a joint letter supporting Anthropic's position. Senators Ed Markey (D-MA) and Chris Van Hollen (D-MD) wrote to Hegseth calling the Pentagon's threats a 'challenge to American innovation.' Anthropic's statement said the supply-chain risk designation was 'legally unsound' and would 'set a dangerous precedent for any American company that negotiates with the government.'