Trump Says a Pentagon Deal With Anthropic Is Possible

President Trump said Monday morning that a deal between the Pentagon and Anthropic is possible — the sharpest reversal yet from an administration that banned it from government contracts four months ago.

"It's possible. We want the smartest people," Trump said on CNBC's Squawk Box. The comment came less than 24 hours after Anthropic CEO Dario Amodei met at the White House with Susie Wiles, Trump's chief of staff, and Treasury Secretary Scott Bessent.

Gregory Barbaccia, the federal chief information officer at the White House Office of Management and Budget, told agency officials this week that the OMB was "setting up protections" to allow government use of a modified version of Anthropic's Mythos model — the successor to Claude — with additional guardrails, Reuters reported. The email was described as routine operational guidance, notable because it described work already underway rather than a new direction. POLITICO reported the emerging framework: Anthropic restores its eligibility for government contracts in exchange for Mythos access restricted to defensive cybersecurity purposes — not the mass surveillance or autonomous weapons uses the company refused when the Pentagon issued its original demand.

The context for that refusal is what made the NSA's position so awkward. The intelligence agency has been running Anthropic's Mythos model despite the DoD blacklist, Axios reported — running it in plain contradiction of the administration's own public posture. One branch of government banned the company. Another never stopped using it.

In February, Defense Secretary Pete Hegseth designated Anthropic a "supply chain risk," blocking every military contractor from using its AI systems. Anthropic refused the Pentagon's demand for unrestricted access to Claude for mass surveillance and autonomous weapons, TechCrunch reported. OpenAI took the contract instead. Anthropic sued in March. A federal judge in San Francisco blocked the ban later that month, ruling the administration had likely retaliated against the company for First Amendment protected speech. A federal appeals court in Washington later declined to extend the block — leaving a split legal posture neither side has resolved.

The diplomatic off-ramp Anthropic gets is precise: it frames the deal as agreeing to defensive cybersecurity use rather than capitulating on its core principles. The administration gets to restore a company it called a national security risk without admitting the ban was porous from the start. Both sides needed a way out of a contradiction that was live.

Amazon's announcement Monday underscores why Anthropic is not desperate for a Pentagon contract. The company said it would invest up to an additional $25 billion in Anthropic, Reuters reported, bringing its total stake to $33 billion, with Anthropic committing to spend more than $100 billion on Amazon Web Services over the next decade. That capital position gives Anthropic leverage in any negotiation — a Pentagon contract is not make-or-break for a company with that balance sheet.

The skeptics are right that the deal is not done. OMB's working group is not a signed contract. The guardrail definitions still have to survive actual negotiation — and if they amount to what Anthropic already offered and the Pentagon rejected, the arrangement could collapse again.

What is notable is where the thaw is coming from. It is not the Pentagon, which remains hostile. It is the intelligence community and the White House — agencies that bypassed the DoD blacklist operationally and now want to formalize what was already happening. The administration called Anthropic a national security risk. Its own spy agency kept using the model. That gap, more than executive goodwill, is what is putting a deal within reach.