Anthropic rejects Pentagon's request for full AI system access

Anthropic CEO Dario Amodei said he “cannot in good conscience accede” to the Pentagon’s request for unrestricted access to its AI systems, citing ethical and safety concerns.

He noted the “Department of War,” not private companies, makes military decisions, but warned that uses like mass surveillance and fully autonomous weapons with no human in the loop could undermine democratic values.

Defense Secretary Pete Hegseth set on February 27 5:01 p.m. deadline for Anthropic to comply or face being labeled a “supply chain risk”—a designation usually reserved for foreign adversaries—or compelled under the Defense Production Act.

Amodei criticized this contradiction, saying one label calls Anthropic a security risk, while another calls its AI essential to national security.

Anthropic is the only frontier AI lab with classified-ready systems for the military, though the DOD is reportedly preparing xAI for similar roles.

Amodei said they want to support the military if their two safeguards are respected and will help transition to another provider if needed.

🔗 Source: TechCrunch

🧠 Food for thought

Implications, context, and why it matters.

A dispute sparked by a covert operation and a poison-pill threat

The standoff followed a January military operation where Anthropic’s Claude was reportedly used to help capture Venezuelan president Nicolás Maduro 1.
An Anthropic employee pressed its partner Palantir, a data analytics and defense software company, on how the AI was used. Palantir then raised the issue with the Pentagon, deepening the fight over Anthropic’s ethical limits 1.
The Pentagon threatened to label Anthropic a “supply chain risk.” That move would block defense contractors from using its products and could derail a planned IPO (initial public offering) 1.
Anthropic does allow some military use. Its policies permit targeted strikes, foreign surveillance, or drone strikes when a human makes the final decision 1.

This battle forces AI labs to choose between principles and survival

The clash could set a precedent on whether any American AI company can keep independent safety policies when government pressure rises 1.
Competitors such as OpenAI and xAI have reportedly accepted the Pentagon’s new terms. Anthropic’s refusal could split the field between “national champions” that comply and outsiders that face blacklisting 2.
Using the Defense Production Act, a U.S. law that can require companies to prioritize government needs, to force removal of software safety limits would have no known prior example and could end up in court 3.
If the government prevails, it could make it easier to push tech firms toward mass surveillance tools or fully autonomous weapons with no human in the loop, reshaping ties between Silicon Valley and the state 1.