OpenAI has disbanded its Preparedness team. The unit was responsible for examining AI models for serious and potentially catastrophic risks, such as the risk that they could help develop biological, chemical, or nuclear threats, or independently compromise other systems, as happened with Hugging Face last month. The Preparedness team is already the third specialized safety team the ChatGPT maker has dissolved.
Responsibilities move into existing departments
According to the Financial Times, the disbanded team's responsibilities are being distributed across existing departments. Dylan Scandinaro, who had only led the Preparedness team since February 2026, will now focus on the safety implications of recursively self-improving AI, meaning systems that use their own capabilities to develop and accelerate subsequent AI generations.
A pattern repeats
The dissolution fits a recognizable pattern. Back in May 2024, OpenAI wound down its Superalignment team after its leader, Jan Leike, left the company. Leike criticized OpenAI at the time for neglecting safety in favor of shiny products. The company subsequently also disbanded its AGI Readiness team, which was meant to assess how well prepared OpenAI and the world are for AI at human intelligence level.
Alongside this restructuring, other safety and ethics executives have left the company. Ethics lead Chloé Bakalar departed in July after less than a year at OpenAI, apparently having been the only person explicitly responsible for ethics at the company. Chief Futurist Josh Achiam and safety chief Johannes Heidecke have also left recently. OpenAI has not yet issued an official public statement about the dissolution.