Pentagon says it has stopped using all Anthropic products
The military's break with Claude is now on the record — and it lands the same week the White House was busy making up with Anthropic's chief executive.
The Pentagon "has ceased the use of Anthropic products," a department official said in a statement to the BBC on Monday — the first on-the-record confirmation that Claude is fully out of the US military's AI stack. The cut had been ordered months ago: defense secretary Pete Hegseth designated Anthropic a national-security supply-chain risk on 27 February and announced the department would stop using it by late August, a deadline that passed with no explanation. What makes the statement new is what it displaces. People familiar with the department's AI use told the BBC that Claude was still being used as recently as last week for research, analysis, intelligence gathering, and even military operations against Iran — largely through Palantir's Maven Smart System, the Pentagon's primary platform for organizing intelligence. The rift was never about capability: the Pentagon pressed Anthropic to strip Claude's safety guardrails for its own uses, and Anthropic refused, citing mass surveillance and fully autonomous weapons.
The messy extraction is half the story. Maven analysts fed satellite imagery and drone footage through Claude to identify potential targets and assemble briefs that moved up the chain of command, which is why the removal took months — Georgetown's Lauren Kahn said the timeline shows such integrations "are not just plug and play," and that once a model is embedded, pulling it out hurts. There is also an unresolved contradiction worth flagging: under secretary Emil Michael said in September that "about 90% has transitioned" and that the Palantir work had moved months earlier, while the BBC's sources say Claude was live in Maven last week. Either the migration was slower than officially reported, or "Maven" means different things to different people — and nobody has reconciled the two. The Pentagon has signed contracts with Google, xAI and OpenAI as replacements, with OpenAI tools spreading faster inside the department. Politically the timing is awkward: Trump met Dario Amodei twice at the White House last week, calling him "great" and "fantastic," while the department was finishing the eviction — Amodei told reporters "how we address those risks is still under discussion" but added that "if we work with the president, we can win safely." We followed the legal side of this fight in The Anthropic–Pentagon ruling isn't a win for one lab. It's a red line for all of them.
What to watch: whether Claude's replacement inside Maven is named explicitly, and whether the warming Trump–Amodei relationship produces a revised designation — or just better optics.
New York City's Council held a first-of-its-kind AI safety hearing on Monday, taking testimony from officials at OpenAI, Anthropic, Google and Meta alongside former Anthropic researcher Jacob Coxon, who warned that the industry is "racing to build and grow our own adversary" and that today's safety fixes may prove "duct tape that will fall off later." The city is weighing legislation that would give New Yorkers a private right of action against AI companies for foreseeable misuse — the bill side of this we covered in New York City wants auditors, a kill switch and a paid whistleblower. The hearing matters because New York is deliberately occupying the regulatory lane Washington keeps leaving open: with Congress stalled, a public record built on sworn testimony is exactly what a city needs before it legislates.
Do you think a vendor should be willing to lose a defense contract over its guardrails? Tell us in the comments.
