OpenAI has reportedly disbanded its preparedness team, according to the Financial Times, in another reorganization of the company's safety work as it moves toward a potential IPO.
The team had been responsible for assessing whether models posed serious risks and developing mitigations. The report says those responsibilities have now been divided into specific areas, including biological and cyber risks, and moved into existing teams. The Verge notes that this follows earlier changes involving OpenAI's AGI readiness and superalignment efforts, as well as departures from safety and ethics roles.
The practical question is not whether risk work still exists, but how independent and visible it remains inside a fast-growing commercial organization. Specialized teams can put expertise closer to product decisions, but breaking up a central preparedness group may make it harder for outsiders to understand who is accountable for cross-cutting model risks.