OpenAI dismantles AI safety preparedness team ahead of IPO
Responsibility for risk assessment divided among existing units, while former team head shifts focus to recursive self-improvement
OpenAI has reportedly disbanded its preparedness team, a unit tasked with assessing serious risks from artificial intelligence models and developing mitigation strategies. According to the Financial Times, the dissolution occurred at the end of last month, marking the latest significant restructuring within the company as it prepares for an anticipated initial public offering.
The team’s specific responsibilities, including oversight of biosecurity and cybersecurity, have been divided among existing teams within the organisation. Dylan Scandinaro, who was appointed head of the preparedness team in February after joining from Anthropic, will now focus on the implications of “recursive self-improving” AI.
This move follows the recent dissolution of OpenAI’s AGI readiness and superalignment teams. The restructuring coincides with the departure of several key safety figures, including ethics lead Chloé Bakalar, Chief Futurist Josh Achiam, and head of safety Johannes Heidecke.
Critics have suggested the company is prioritising product development over safety. Jan Leike, who resigned from OpenAI in 2024, told the Financial Times that the company was ignoring safety in favour of creating “shiny products.” The preparedness team’s original mandate included assessing risks such as models potentially going rogue or hacking other companies, though such characterisations simplify the broader scope of risk assessment.
The disbanding of the preparedness team is part of a broader upheaval at OpenAI over recent years. As the company heads towards a massive IPO, the shift in safety structures has drawn attention from investors and industry observers monitoring how the firm balances rapid product development with rigorous risk management.
