Anthropic may regain Pentagon access as Trump signals possible deal

US President Donald Trump said Tuesday that a deal allowing Anthropic's artificial intelligence models to be used by the Pentagon is possible, signaling a potential thaw in the administration's dispute with the AI startup.

Trump told CNBC that Anthropic officials recently visited the White House and held "very good talks" with his administration.

"They're shaping up," said Trump, adding that the company could be "of great use" to the Pentagon.

The remarks mark a softer tone after the Pentagon labeled Anthropic a supply chain risk in March following a breakdown in negotiations, prompting defense contractors to certify they were not using the company's Claude AI models in military-related work.

Trump also ordered federal agencies to immediately stop using Anthropic's technology, though that directive was later temporarily blocked by a federal judge.

Tensions appear to have eased in recent weeks after Anthropic CEO Dario Amodei met senior officials Friday at the White House to discuss the company's newly launched Mythos model, which has drawn attention for its advanced cybersecurity capabilities.

Anthropic said it has been engaged in discussions with government officials about Mythos, which was rolled out earlier this month to a limited group of companies.

The company's relations with Washington deteriorated after negotiations on the deployment of Claude on the Pentagon's GenAI.mil platform collapsed last year.

The Pentagon sought broad access to Anthropic's models for all lawful purposes, while the company wanted guarantees that its technology would not be used for fully autonomous weapons or domestic mass surveillance.

Founded in 2021, Anthropic signed a $200 million Pentagon contract in July, but the dispute led to its effective blacklisting from parts of the federal system.