Pentagon sets Friday deadline for Anthropic to drop AI ethics rules

Defense Secretary Pete Hegseth has given Anthropic until 5:01 p.m. Friday to remove restrictions on how its AI can be used by the U.S. military, or face contract termination and designation as a supply chain risk. In a Tuesday meeting with Anthropic CEO Dario Amodei, Hegseth warned the company to set aside its usage policy limits or have the Cold War-era Defense Production Act invoked against it -- a law that gives the federal government broad authority to compel private companies to act in the national interest.

Anthropic maintains two specific restrictions it refuses to lift: its AI cannot be used for mass domestic surveillance of U.S. citizens, and it cannot make autonomous targeting decisions in physical attacks without human oversight. These are not informal preferences -- they are published usage policies that Amodei reaffirmed directly to Hegseth on Tuesday. The dispute escalated after a report surfaced that Anthropic's Claude model was used by the U.S. military during a January operation to capture former Venezuelan President Nicolas Maduro, putting the company's Pentagon contracts under public scrutiny. The standoff is significant because it is one of the first direct, documented cases of a U.S. government agency using legal and contractual pressure to force a major AI developer to weaken its own safety guardrails. If the Pentagon succeeds, it sets a precedent that national security priorities can override an AI company's self-imposed ethical limits. If Anthropic holds its position, it risks losing a major government customer and may face legal action under the Defense Production Act.