OpenAI employees reportedly warned leaders about security risks for years

OpenAI has since taken steps to address these issues. The company stated that sensitive work is now held in separate repositories with strict access controls, and the main monorepo no longer contains critical model details like weights. Under Chief Information Security Officer Dane Stuckey, who joined in October 2024, OpenAI now conducts regular security reviews of internal tooling. Senior executive Thibault Sottiaux has also led an initiative to secure internal systems, reassigning a quarter of production engineers to security work in August 2026.

Additionally, OpenAI announced plans for another internal security sprint with Cathedral, where researchers will use OpenAI models to challenge defenses. The company is more than quadrupling rewards for critical findings in its bug-bounty program. This follows the October 2026 arrest of engineer Abdulrahman Alfozan, who was charged with using a recording device in an executive conference room. Alfozan has been charged, not convicted. The company faces the challenge of balancing its open research culture with the need for stricter security as AI systems become more capable.