TL;DR
Anthropic CEO Dario Amodei has returned to negotiations with the Pentagon after talks collapsed last week. He is reportedly discussing a new contract with Under-Secretary of Defense Emil Michael. Meanwhile, a leaked internal memo from Amodei to staff described OpenAI’s Pentagon deal as “safety theatre” and attributed the breakdown to the fact that Anthropic has not donated to or praised President Trump.
New Negotiations
According to the Financial Times, Amodei is in discussions with Emil Michael about a contract that would allow the US military to continue using Claude. This comes after Michael publicly attacked Amodei on social media, calling him a “liar” with a “God-complex” who was “putting our nation’s safety at risk.”
The stakes are high. Defence Secretary Pete Hegseth’s supply-chain risk designation would force firms across the US tech ecosystem to drop Claude if they want to keep working on defence contracts. Until last week, Claude was the only AI system with security clearance to handle classified information, and reports indicate it was actively used in both the Venezuela raid and the Iran strikes.
The Leaked Memo
A leaked internal memo from Amodei to Anthropic staff has added fuel to the dispute. In it, Amodei reportedly described messaging from both the Pentagon and OpenAI as “straight up lies” and called OpenAI’s military deal “safety theatre.”
He also suggested the relationship soured because, unlike OpenAI and its executives, “we haven’t donated to Trump” and “we haven’t given dictator-style praise to Trump.” OpenAI CEO Sam Altman has been among the Silicon Valley leaders building relationships with the administration, and co-founder Greg Brockman is a Trump mega-donor.
Amodei also revealed that the Pentagon had been close to accepting Anthropic’s terms but asked the company to delete a specific phrase about “analysis of bulk acquired data” — the exact scenario Anthropic was most concerned about. “We found that very suspicious,” he wrote.
Looking Forward
The negotiations are complicated by the fact that Claude remains actively embedded in classified military operations even as the administration threatens to ban it. For the broader AI industry, the outcome will signal whether companies can maintain safety boundaries with government clients, or whether political alignment has become a prerequisite for federal contracts.