OpenAI Disbands Preparedness Team, Splits Work Across Groups
OpenAI dissolved its preparedness team at the end of July 2026, reassigning responsibility for assessing model risks to separate domain-specific units, according to a Financial Times report.
👁 118 reads
OpenAI dissolved its preparedness team at the end of July 2026, reassigning its model-risk work to existing domain-focused groups rather than maintaining a dedicated unit, according to a report citing the Financial Times.
- OpenAI disbanded its preparedness team at the end of July 2026, with duties split across existing teams by risk area such as bio and cyber
- Team head Dylan Scandinaro, hired from Anthropic in February 2026, will now focus specifically on recursive self-improving AI
- The move follows the earlier dissolution of OpenAI's AGI readiness and superalignment teams, and the recent departures of its ethics lead, chief futurist, and head of safety
The preparedness team had been responsible for evaluating whether OpenAI’s models posed serious risks - including the possibility of a model autonomously hacking external systems - and for developing mitigations. Under the restructuring, that remit has been divided by subject area, such as biosecurity and cybersecurity, and absorbed into teams that already exist inside the company.
Dylan Scandinaro, who led the preparedness team after being recruited from Anthropic in February 2026, will not be departing. His new focus will be on the implications of what the company calls “recursive self-improving” AI, a scenario in which a model iteratively enhances its own capabilities.
Google · Your sourcesQuantslant higher in your Google - one clickAdd to preferred sources →A pattern of safety team reductions
The disbanding is the latest in a series of structural changes to OpenAI’s safety apparatus. The company has previously wound down both its AGI readiness team and its superalignment team, the latter of which had been co-led by Ilya Sutskever before his departure. More recently, three prominent figures left the organisation: ethics lead Chloé Bakalar, chief futurist Josh Achiam, and head of safety Johannes Heidecke.
Critics have pointed to the cumulative effect of those changes as a signal of shifting priorities. Jan Leike, who resigned from OpenAI in 2024, told the Financial Times that the company was ignoring safety in favour of creating “shiny products.”
OpenAI is navigating all of this while preparing for what is widely expected to be a large initial public offering. The company has also been restructuring its broader corporate form - converting from a capped-profit entity toward a more conventional for-profit structure - a process that has coincided with repeated turnover in its safety and ethics leadership.
The preparedness function itself was introduced in late 2023 as a formal mechanism for evaluating catastrophic and frontier risks before models were deployed. Fragmenting it into existing operational teams removes the dedicated organisational unit, though the company has indicated the underlying work will continue within those groups.
More on this from QUANTSLANT
magazine·Stripe reportedly acquires AI model gateway OpenRouter for over $7BStripe has finalized a deal to acquire AI gateway startup OpenRouter at a price above $7 billion, according to a Bloomberg report published August 16, 2026.