Anthropic Vows Legal Fight Against Pentagon Sanction in AI Feud

Bloomberg reports that Anthropic has pledged to contest in court a Pentagon decision to label it a risk to the US supply chain, using an authority normally aimed at foreign adversaries. Chief Executive Dario Amodei said in a blog post that the company considers the action legally unsound and sees no alternative to a court challenge. Defense officials had notified the company of it a day earlier.

The report links the move to weeks of tense talks over military use of Anthropic's technology. Discussions collapsed after the company asked for assurances that its AI would not support mass surveillance of Americans or autonomous weapons, and Defense Secretary Pete Hegseth then threatened the designation. The step endangers a $200 million contract for classified AI tools. Amodei said the cited statute, section 3252, is narrow enough to leave unrelated business untouched, and Microsoft said it can keep working with Anthropic on non-defense projects. The company must nonetheless stop working with Palantir, including on a platform used in the Iran campaign. Hegseth wrote to congressional leaders that the firm's restrictions on use create national security risks. After President Trump demanded that the government drop the firm, the Treasury Department and the General Services Administration said they will stop doing business with it.

Charlie Bullock of the Institute for Law & AI called the label unprecedented for an American company and beyond what the law was meant for. A defense official said the determination was effective immediately, yet Claude is reportedly still in use for Iran operations, and Hegseth had outlined a six-month transition. Amodei described recent talks as productive, but Emil Michael, an under secretary of defense, said no negotiation is active. Amodei also apologized for the tone of a leaked internal memo about OpenAI.