Anthropic CEO Dario Amodei stirred things up this week with a sharp internal memo to his staff. He called out rival OpenAI and its boss Sam Altman over a fresh deal with the US Department of Defense, now called the Department of War under the Trump administration. The memo leaked fast and grabbed headlines. Amodei slammed OpenAI’s Pentagon agreement as mostly show, not real protection against risky uses like mass surveillance on Americans or fully autonomous weapons.
The drama started when talks between Anthropic and the Pentagon broke down last week. The government wanted loose rules for using Anthropic’s AI model Claude. Anthropic pushed back hard on its red lines. Now, fresh reports say Amodei has restarted chats with defence officials. This comes right after his tough words in the memo.
Amodei’s strong words on OpenAI
As per The Information, In the memo, Amodei went after OpenAI’s messaging. He wrote, “I want to be very clear on the messaging that is coming from OpenAI, and the mendacious nature of it.” He called their safety setup “maybe 20% real and 80% safety theater.” Things like model refusals or extra software layers don’t hold up well in military settings, he argued. A model can’t really tell if data comes from US citizens or if a human stays in control for weapons.
Amodei pointed out jailbreaks happen easy. He said refusals fail often. He added that monitoring by employees only works in a small number of cases. “The main reason OAI accepted them and we did not is that they cared about placating employees, and we actually cared about preventing abuses,” he wrote.
He also hit back at claims that OpenAI’s terms match what Anthropic turned down. “It is false that ‘OpenAI’s terms were offered to us and we rejected them’,” he said. At the same time, he called OpenAI’s protections weak against the big worries.
Why Anthropic clashed with the Government
Amodei explained the bad blood with the Trump team. Anthropic stood firm on safety. They refused to bend. He wrote, “We haven’t given dictator-style praise to Trump (while Sam has), we have supported AI regulation which is against their agenda, we’ve told the truth about a number of AI policy issues (like job displacement), and we’ve actually held our red lines with integrity rather than colluding with them to produce ‘safety theater’.”
The Pentagon pushed for “all lawful use” without tight limits. Near the end, they offered to take Anthropic’s terms if one phrase got cut. “Near the end of the negotiation the department offered to accept our current terms if we deleted a specific phrase about ‘analysis of bulk acquired data,’ which was the single line in the contract that exactly matched the scenario we were most worried about. We found that very suspicious,” Amodei noted.
On autonomous weapons, he said current policy needs a human in the loop. But that can change fast. “A lot of OpenAI and DoW messaging just straight up lies about these issues or tries to confuse them,” he added.
Talks restart amid the tension
According to the Financial Times, Amodei now talks again with Emil Michael, the under-secretary of defense for research and engineering. The goal? Fix a contract so the military keeps using Anthropic tech. It could dodge the supply chain risk label that Pete Hegseth threatened. Talks collapsed last week after fights over surveillance language. Michael even called Amodei a “liar” with a “God complex” in one heated moment.
Anthropic had a $200 million deal from last July. It made Claude the first big AI in classified US settings. The company wants to stay in defense work but not drop its safeguards. Resuming dialogue hints at a possible compromise. Or it could drag the feud on longer.