OpenAI restructures safety operations by dissolving preparedness team

OpenAI has disbanded its internal preparedness team at the end of last month, marking another significant shift in the company's safety governance structure. The team was originally tasked with evaluating advanced models for severe risks, including potential catastrophic harms, and developing corresponding mitigation strategies.

Redistribution of safety oversight

Rather than maintaining a standalone unit, responsibility for specific risk categories such as biological and cyber threats has now been divided and integrated into existing engineering and product teams. Dylan Scandinaro, who previously led the preparedness team after joining from Anthropic in February, will transition to focus specifically on the implications of recursive self-improving artificial intelligence.

Broader organizational changes

This structural adjustment follows a series of departures among senior safety and research personnel over recent years. Previous dissolutions included the AGI readiness and superalignment teams. Recent exits from the organization include ethics lead Chloé Bakalar, chief futurist Josh Achiam, and head of safety Johannes Heidecke.

Critics of the trajectory, including former researchers who departed the organization previously, have raised concerns regarding the prioritization of commercial product development relative to long-term safety research. As the organization prepares for a potential public offering, these internal realignments reflect an ongoing evolution in how artificial intelligence safety and risk assessment are managed at scale.