OpenAI reportedly disbands its AI safety preparedness team
The Financial Times reports OpenAI dissolved the team tasked with assessing serious AI model risks at the end of last month. It's the latest in a string of safety-related departures and team cuts at the company.
OpenAI has disbanded its preparedness team, the group responsible for assessing whether AI models pose serious risks and developing ways to mitigate them, the Financial Times reported. The team's mandate included guarding against scenarios such as a model going rogue and hacking outside systems. According to FT, that responsibility has now been split among specific risk areas, such as biological and cyber threats, and folded into existing teams rather than kept as a standalone unit.
Part of a broader shakeup
The move is the latest change at OpenAI, which has seen considerable internal upheaval as it moves toward an anticipated massive IPO. Over the past several years, the company has gradually dismantled its more research-focused safety structure, previously dissolving its AGI readiness team and its superalignment team.
Several safety- and ethics-focused staff have also recently departed, including ethics lead Chloé Bakalar, chief futurist Josh Achiam, and head of safety Johannes Heidecke. The departures have prompted critics, including Jan Leike, who resigned from OpenAI in 2024, to tell FT that the company is deprioritizing safety in favor of building "shiny products."
Dylan Scandinaro, who led the preparedness team after joining OpenAI from Anthropic in February, will now turn his focus to the implications of "recursive self-improving" AI.
