Anthropic to challenge Pentagon’s ‘supply chain risk’ label in court; CEO apologises for tone of leaked memo

Anthropic will challenge in court a decision by the US Department of War to designate the AI firm as a national security “supply chain risk,” CEO Dario Amodei announced on Friday.

Amodei said that the company received a letter from the department confirming the designation, which he argued was legally unsound and narrowly applicable.

“We do not believe this action is legally sound, and we see no choice but to challenge it in court,” Amodei said.

The designation means Anthropic’s AI tools, including its model Claude, could face restrictions in contracts tied directly to the Pentagon.

Amodei said the measure applies only when Claude is used as part of specific Department of War contracts, and does not broadly restrict customers who also work with the government.

“The vast majority of our customers are unaffected,” Amodei insisted.

The dispute comes amid growing competition among major US AI companies to supply advanced tools to the military and intelligence community.

Anthropic’s models have been widely deployed across the Department of War and national security agencies for tasks including intelligence analysis, modeling and simulation, operational planning and cyber operations.

But tensions escalated after Anthropic refused to remove certain safeguards on its AI models despite pressure from the Pentagon.

The department later signed a deal with rival AI firm OpenAI to integrate its technology into classified military systems.

OpenAI CEO Sam Altman announced the agreement late last week, saying the company’s tools would be used in defense networks while maintaining safeguards similar to those Anthropic had proposed.

Days later, Altman said the company would make further changes, including ensuring its systems would not be “intentionally used for domestic surveillance of US persons and nationals.”

He also said intelligence agencies such as the National Security Agency would require additional contract modifications before using the technology.

Altman acknowledged the rollout of the deal had been rushed.

“The issues are super complex, and demand clear communication,” he wrote on social platform X. “We were genuinely trying to de-escalate things and avoid a much worse outcome.”

Anthropic, however, has drawn a harder line on how its technology can be used. Last week the company said it would not remove safeguards preventing its AI models from being used for fully autonomous weapons or mass domestic surveillance.

Amodei said the company supports US national security but would not agree to allow “any lawful use” of its technology without restrictions.

“I believe deeply in the existential importance of using AI to defend the United States and other democracies,” he wrote. “But we cannot in good conscience accede” to demands that remove key guardrails.

In a memo to staff reported by tech publication The Information, Amodei also criticised OpenAI’s approach to military partnerships, calling it “safety theater” and accusing the company of misleading messaging around its Pentagon deal.

Amodei later apologised for the tone of an internal post that was leaked to the media, saying it had been written during a turbulent day that included government announcements about Anthropic’s removal from federal systems and news of OpenAI’s defence contract.

Amodei said Anthropic would continue supporting US national security operations during any transition period.

“Our most important priority right now is making sure that our warfighters and national security experts are not deprived of important tools in the middle of major combat operations,” he said.