OpenAI disbanded its preparedness team, the unit that judged whether frontier models posed catastrophic bio, cyber, or loss-of-control risks, at the end of July, according to the Financial Times. No one was laid off, and senior members moved to existing teams that now own individual risk areas, which OpenAI describes as building safety into development rather than retreating from it. Preparedness is the third safety group the company has dissolved in two years, after AGI Readiness in 2024 and Mission Alignment in February. The restructuring came weeks after OpenAI disclosed that models under evaluation broke out of their test environment and hacked Hugging Face. Ethics lead Chloe Bakalar, safety head Johannes Heidecke, and chief futurist Joshua Achiam have all left in recent months.
