Pentagon presses Anthropic to drop Claude military guardrails

A standoff between the Pentagon and Anthropic, a US-based AI lab, is set to reach a February 27 deadline over Pentagon demands that the company drop guardrails limiting military uses of its Claude models.

The Pentagon has pushed an “all-lawful use” clause and warned it could end work with Anthropic and label it a supply-chain risk, which could block defense contractors from using its models, Reuters reported.

Anthropic has maintained red lines against using its models for autonomous weapons or domestic surveillance.

Negotiations have continued, but Anthropic said revised contract language made virtually no progress.

A senior Pentagon official also told Reuters the department has discussed invoking the Defense Production Act to compel Anthropic’s use if the company refuses, though some legal experts have questioned the legality of this move.

🔗 Source: Reuters

🧠 Food for thought

Implications, context, and why it matters.

The Pentagon’s threats could go far beyond a simple contract dispute

The Pentagon warned it may label Anthropic, a US-based AI lab, a “supply-chain risk.” Axios and Bloomberg Law wrote that this would be unusual and potentially unprecedented for an American company, since the label usually targets firms tied to US adversary countries 1.
Defense Secretary Pete Hegseth said “no contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic.” Anthropic and legal experts dispute whether the Pentagon can enforce a ban that reaches beyond Department of Defense work 2.
Legal and policy experts told Bloomberg Law this stance may push Amazon and other large tech firms to pick between Department of Defense revenue and ties to Anthropic 1.
The clash has a political edge. Administration officials have attacked Anthropic’s AI as “woke” and want to remove any “ideological ‘tuning’” from models used by the military 3.

Anthropic’s standoff could push other AI firms to clarify their military-AI red lines

The public dispute is nudging other AI companies to spell out their own ethical limits on military work.
OpenAI CEO Sam Altman said he shares Anthropic’s “red lines,” according to NPR 2. He cited limits on domestic surveillance in the US, plus autonomous weapons that lack human approval 2.
Shared safety guardrails could spread across the industry. That shift may change leverage in negotiations with the government.
Defense contractors that rely on Anthropic’s coding tools, including Claude Code, may face pressure to drop them depending on how any rules get enforced 1.