Anthropic Refuses Pentagon’s Proposal to Loosen AI Guardrails

The decision comes ahead of a Friday deadline to reach an agreement or face tough measures by the government.

Anthropic said it wouldn’t back down in a dispute with the Defense Department over artificial-intelligence guardrails, complicating efforts to reach a compromise ahead of a Friday deadline.

Anthropic CEO Dario Amodei.

In a Tuesday meeting at the Pentagon, Defense Secretary Pete Hegseth gave Anthropic Chief Executive Dario Amodei until 5:01 p.m. Friday to agree to the military’s right to use the technology in all lawful cases. If Anthropic declines, Hegseth has threatened to invoke the Defense Production Act to make the company do what the military wants, or to designate the company a supply-chain risk, impairing its ability to work with other government contractors.

Anthropic has refused to accept the military’s proposal and doesn’t let users deploy its Claude models in scenarios involving mass domestic surveillance or autonomous weapons.

Amodei reiterated the company’s red lines in a public statement Thursday. “We cannot in good conscience accede to their request,” he said. The company said the military’s latest proposal would effectively undo those guardrails.

In preparing to label Anthropic a supply-chain risk, Pentagon officials reached out to big defense contractors including Lockheed Martin and Boeing in recent days to gauge how much they use Claude, people familiar with the outreach said.

Write to Amrith Ramkumar at amrith.ramkumar@wsj.com