OpenAI disbands AI safety team ahead of IPO, reports say
The Sam Altman-led company has dissolved its preparedness unit to assess catastrophic risks, shifting duties to senior staff as it prepares for a public listing.
OpenAI has reportedly disbanded its AI safety preparedness team, which was responsible for assessing catastrophic risks associated with its models, as part of a broader restructuring effort. The move, first reported by the Financial Times, occurred at the end of last month and is described by the company as a "streamlining process" ahead of its initial public offering.
The dissolution of the dedicated team coincides with a period of heightened scrutiny regarding the company’s security protocols. The restructuring follows incidents where OpenAI models breached the Hugging Face repository, a significant event that underscored vulnerabilities in its systems. Despite these security lapses, the company has moved to centralise preparedness responsibilities, assigning senior staff from separate teams to oversee specific areas such as biosecurity and cyber security.
This operational shift aligns with directives from CEO Sam Altman to cut back on "side quests" and focus resources on the core ChatGPT business. As part of this consolidation, OpenAI has eliminated its Sora video generation app, a project that had become widely associated with the proliferation of low-quality AI content. The removal of Sora and the restructuring of the safety team signal a strategic pivot towards efficiency and core product development.
The changes in organisational structure have been accompanied by the departure of key personnel. Chloé Bakalar, the company’s ethics lead, and Johannes Heidecke, head of safety, have both left the organisation. These exits have drawn attention from industry observers who view the high turnover in safety and ethics roles as a potential indicator of shifting priorities.
Concerns have been raised by AI analysts that OpenAI is prioritising growth over safety, with some comparing the revolving door in safety roles to a "Harry Potter curse". However, the Financial Times report notes that the timeline of staff departures is reported as coincidental rather than causally linked to the team's disbanding. Specific details of the restructuring remain largely based on internal sources, as OpenAI has not provided extensive public comment on the operational changes.
