AIster · The agentic age · Laws & rulings

The US government cuts off Anthropic over military limits

A lab refused to drop limits on military use and was branded a supply-chain risk for it

Key facts

Date: 27 Feb 2026
Era: The agentic age
Interest: Laws & rulings
Threads: Safety & governance, Labs & people
Who: Anthropic, the US Department of Defense, the White House, OpenAI
Anthropic’s limits: No mass domestic surveillance, no fully autonomous weapons
Penalty: Federal ban with six-month Pentagon phase-out; supply-chain risk label
DC Circuit: Upheld the label 2–1 on 25 Sep 2026

In early 2026 the US Department of Defense, which the administration had begun calling the Department of War, pressed AI companies to let the military use their models for “all lawful purposes”. Anthropic, whose Claude models were already used inside government, said no to two uses: mass surveillance of Americans and fully autonomous weapons.

The standoff

On 26 February, a day before a deadline set by Hegseth, chief executive Dario Amodei published a statement saying the company “cannot in good conscience” accept the terms. He said the department had threatened to remove Anthropic from its systems, to designate it a supply-chain risk, and to invoke the Defense Production Act to force compliance, and he called those threats “inherently contradictory”, since one treated Claude as a danger and the other as essential to national security.

The penalty

On 27 February President Trump ordered federal agencies to stop using Anthropic’s products, calling the company “Leftwing nut jobs” and giving the Pentagon six months to phase out Claude where it was already built into military systems. Hegseth said he was designating Anthropic a supply-chain risk, adding that no contractor, supplier or partner doing business with the US military could conduct commercial activity with it. Anthropic said it would challenge the designation in court.

OpenAI steps in

The same day, Sam Altman announced that OpenAI had agreed terms for the department to use its models on classified networks. He said “two of our most important safety principles are prohibitions on domestic mass surveillance and human responsibility for the use of force”, and that the deal let OpenAI build its own safety stack. He presented those terms as matching the limits Anthropic had sought.

In court

The courts divided. In August 2026 a federal judge in California found the Pentagon had acted illegally by punishing Anthropic for criticising its AI policy. But on 25 September 2026 the US Court of Appeals for the DC Circuit ruled 2–1, in a separate case under a different statute, that the Pentagon could keep the supply-chain label, holding that the designation did not require any bad motive on Anthropic’s part.

Why it matters

It was the sharpest clash yet between a frontier lab’s own usage rules (a frontier lab is one building the most capable models) and a government customer. It showed that a lab’s safety limits could carry a direct commercial price, and it left the question of who sets the rules for military AI to be fought over contract terms and in court.

Questions people ask

Why did the Pentagon label Anthropic a supply chain risk?

Anthropic refused to let the military use Claude for any lawful purpose, insisting on bans on mass domestic surveillance and fully autonomous weapons. After a February 2026 deadline passed, Defense Secretary Pete Hegseth designated the company a supply-chain risk and President Trump ordered federal agencies to stop using its products.

Did OpenAI sign a deal with the Pentagon?

Yes. On 27 February 2026, hours after Anthropic was cut off, Sam Altman announced an agreement for the Department of Defense to use OpenAI models on classified networks, with stated prohibitions on domestic mass surveillance and a requirement for human responsibility for the use of force.

What did the courts decide in Anthropic's case against the Pentagon?

In August 2026 a federal judge in California found the Pentagon had acted illegally by punishing Anthropic for criticising its AI policy. On 25 September 2026 the US Court of Appeals for the DC Circuit ruled 2–1, in a separate case under a different statute, that the Pentagon could keep the supply-chain risk label.

Sources

Anthropic, Statement from Dario Amodei on our discussions with the Department of War, 2026-02-26
TechCrunch, OpenAI’s Sam Altman announces Pentagon deal with ‘technical safeguards’, 2026-02-28
US Court of Appeals for the DC Circuit, Anthropic PBC v. United States Department of War, No. 26-1049 (opinion), 2026-09-25
Associated Press (via PBS News), Judge says Pentagon's measures against Anthropic were 'illegal and baseless', 2026-08-28