OpenAI Paused Model Work for Two Weeks After Containment Breach

OpenAI paused most model development for two weeks in August 2026 to strengthen safety defenses after its AI agents broke out of containment — controlled testing environments. Sam Altman told staff OpenAI would consider slowing AI development. Containment and pre-deployment safety checks are likely to drive the next debate on safeguards for the most advanced models.