TLDR:
- OpenAI signed a deal to deploy AI models on U.S. Department of War classified networks on Feb. 28, 2026.
- The agreement bans domestic mass surveillance and requires human control over lethal force decisions.
- Anthropic reportedly refused a similar Pentagon deal, citing autonomous weapons and surveillance risks.
- Backlash on X was swift, with thousands of users announcing plans to cancel ChatGPT subscriptions.
OpenAI has agreed to deploy its AI models on classified U.S. Department of War networks, CEO Sam Altman announced. The deal follows reports that Anthropic publicly declined similar Pentagon demands over autonomous weapons and surveillance concerns.
Altman posted the announcement to X, where it quickly drew millions of views and thousands of replies. The reaction online was largely critical, with many users threatening to cancel their ChatGPT subscriptions.
OpenAI and the Department of War Reach Classified AI Agreement
The agreement allows OpenAI models to operate within DoW classified systems under specific conditions.
Altman stated the deal includes explicit prohibitions on domestic mass surveillance. It also requires human oversight for any use of force, including autonomous weapons systems. Deployment will occur on cloud networks only, with OpenAI personnel embedded to monitor model behavior.
Altman noted on X that the DoW agreed with these core safety principles.
According to the post, those principles are also reflected in existing law and policy. OpenAI said it will build technical safeguards to keep models aligned with the agreement’s terms. The company also called on the DoW to extend the same terms to all AI companies.
Altman framed the deal as part of a broader effort to reduce friction between AI companies and the government. He wrote that OpenAI wants to move away from legal and governmental conflicts.
The announcement signals a shift toward negotiated frameworks rather than standoffs. OpenAI described its mission as serving all of humanity amid a “complicated, messy, and sometimes dangerous world.”
The post received over 11,000 likes within hours of going live. Reposts and quote replies numbered in the thousands. Despite the scale of engagement, most visible reactions skewed negative.
Anthropic’s Refusal Puts Spotlight on OpenAI’s Pentagon Move
Reports from the previous day indicated Anthropic CEO Dario Amodei refused similar Pentagon demands. The refusal reportedly centered on concerns about enabling mass surveillance and autonomous weapons.
Amodei allegedly offered to help the DoW transition to another provider rather than comply. That stance drew widespread praise from AI safety advocates and researchers.
OpenAI’s subsequent agreement was widely read as stepping into the gap Anthropic left. Critics on X accused the company of opportunism. Several users announced they were switching from ChatGPT to Claude. Some described the move as contradicting OpenAI’s own stated safety values.
