New YorkLate city edition1098 stories
Info Super Site
Technology

Anthropic rejects a Pentagon request as OpenAI agrees to deploy

27 Feb 2026, 12:30 pm

Anthropic chief executive Dario Amodei rejected a United States Department of Defense request to loosen security safeguards on the Claude model. The request concerned potential use in mass surveillance and autonomous weapons systems. On the same stretch of days, OpenAI announced an agreement to deploy its technology on the department's network. The two decisions pointed in opposite directions.

A large language model is a software system trained to generate and analyze text. Claude is Anthropic's model. Safeguards are limits the company builds in so the system is not used for certain purposes. Amodei refused to loosen those limits for the military request as it was described.

Mass surveillance is the monitoring of large numbers of people. Autonomous weapons are weapon systems that can select or engage targets with reduced human control. Those were the uses named in the request Anthropic rejected. The report did not say the Pentagon had already deployed Claude for either purpose. It said the company declined to relax safeguards to allow that potential use.

OpenAI's announcement was a separate corporate decision. The company said it had reached an agreement with the Department of Defense to deploy its technology on the department's network. The report noted the deal in an account that also referred to ethical safeguards, but it did not list the terms. Deployment on a network means the software would run inside the department's systems under that agreement.

The two firms are both builders of advanced artificial-intelligence systems, and the department is the same customer. One chief executive refused a change to safeguards tied to surveillance and autonomous weapons. The other company said it had a deployment agreement. The report did not claim the agreements covered identical tasks.

Amodei turned down a Defense Department request to weaken Claude's safeguards for mass surveillance and autonomous weapons. OpenAI said it had agreed to deploy its technology on the department's network. The refusals and the agreement were the technology developments reported. No contract values were disclosed.

More in Technology