OpenAI revises US classified-AI deal after public backlash
Mar 3rd 2026
OpenAI says it is changing a Pentagon agreement after users and staff protested, adding limits on surveillance and extra approvals for intelligence use while critics warn about AI safety in military settings.
- OpenAI said it amended its Pentagon contract to add guardrails for classified AI deployments.
- CEO Sam Altman acknowledged the announcement was rushed and said the company will bar intentional domestic surveillance of US persons.
- Intelligence agencies including the NSA will need a follow-on contract modification before using OpenAI systems.
- OpenAI faced strong user backlash and a reported 295% surge in ChatGPT mobile app uninstalls over the weekend.
- Anthropic rose in App Store rankings after OpenAI's announcement and was previously blacklisted by the Trump administration over its refusal to drop a ban on fully autonomous weapons.
- Experts warn the departure of safety-focused firms from US defence contracts raises concerns, while militaries say AI is used for logistics and analysis but kept with human oversight due to risks like model hallucinations.