The Pentagon wants an AI vendor that will sign first and argue later. Anthropic is doing the opposite, and the standoff is getting a deadline, a contract threat, and a cameo from the Defense Production Act.
What You Should Know
Anthropic CEO Dario Amodei says the company will not accept Defense Department terms that allow unrestricted use of its Claude model. The Pentagon says it wants only lawful uses, and it has threatened to pull Anthropic’s contract by February 27th, 2026.
The unusually public clash, reported by The Associated Press and carried by PBS NewsHour, puts Defense Secretary Pete Hegseth, Pentagon spokesman Sean Parnell, and Silicon Valley’s safety-first pitch into the same pressure cooker.
The Ultimatum and the Leverage
According to AP, Hegseth delivered an ultimatum after meeting with Amodei: open Anthropic’s tech for unrestricted military use or risk losing the deal. Military officials also raised sharper tools, including branding Anthropic a supply chain risk or using the Defense Production Act to expand authority.
Amodei called that two-track threat internally inconsistent, writing that “those latter two threats are inherently contradictory: one labels us a security risk; the other labels Claude as essential to national security.” If the Pentagon walks, he said, Anthropic would work on a transition to another provider.
That matters because the Pentagon already does business with other AI heavyweights. AP reported that Anthropic is the last of its peers, including Google, OpenAI, and Elon Musk’s xAI, not supplying its technology to a new internal US military network.
The Guardrails Anthropic Refuses To Remove
Anthropic says its sticking point is not whether to work with the military, but how. The company said new Defense Department language made little progress toward blocking Claude from being used for mass surveillance of Americans or in fully autonomous weapons.
Parnell pushed back publicly, saying the military has no interest in mass surveillance of Americans and does not want autonomous weapons operating without human involvement, according to AP’s account of his social media statements. He also said the Pentagon wants Anthropic’s model for lawful purposes and argued that wider access would keep companies from jeopardizing operations.
Why the Lawyers Suddenly Look Like the Plot
The dispute is also playing out amid a broader culture war inside the Pentagon over who slows operational decisions. AP noted that Hegseth told Fox News in February 2025 that the department wants lawyers who give constitutional advice and do not exist to be roadblocks, and AP reported he later fired the top lawyers for the Army and the Air Force without explanation.
On Capitol Hill, the public nature of the fight is becoming a scandal in itself. Sen. Thom Tillis, a North Carolina Republican, summed up the awkwardness with a blunt question: “Why in the hell are we having this discussion in public?” Sen. Mark Warner, the top Democrat on the Senate Intelligence Committee, said he was disturbed by reports that the Pentagon was trying to bully a leading US company.
What to watch next is simple and expensive: whether the Defense Department backs off its wording, or whether it cuts Anthropic loose and treats a safety-minded refusal as a procurement problem instead of a policy warning.
