OpenAI dissolves Preparedness team that evaluated catastrophic AI risks
OpenAI has shut down its Preparedness team, the internal unit charged with evaluating whether the company's own AI models could pose catastrophic risks, and reassigned that work to existing groups rather than maintaining a dedicated unit. Several safety staffers have left following the change. Reporting summarized in the evidence packet states the dissolution has triggered internal unrest and staff departures as catastrophic-risk evaluation shifts from a specialized Preparedness function to distributed oversight across existing teams. For builders and operators who depend on frontier-model APIs, the move reduces public visibility into how one of the largest labs structures pre-deployment catastrophic-risk review. The packet does not specify which groups now own Preparedness responsibilities or publish new evaluation criteria, leaving the post-reorganization safety posture incompletely described in public reporting.
OpenAI dissolves Preparedness team that evaluated catastrophic AI risks
OpenAI shut down its "Preparedness" team, which evaluated whether the company's own AI models could pose catastrophic risks. The work has been parceled out to existing groups, and several safety staffers have left.
Key takeaway
OpenAI's dismantling of its dedicated Preparedness unit means catastrophic-risk evaluation is now embedded across existing teams without a single accountable safety organization.
What happened
OpenAI shut down its Preparedness team, which evaluated whether the company's own AI models could pose catastrophic risks, according to The Decoder.
The work has been parceled out to existing groups, and several safety staffers have left following the dissolution of the specialized unit tasked with catastrophic-risk evaluation.
Evidence
OpenAI shut down its Preparedness team, which evaluated catastrophic risks from its own AI models.
The Decoder · attributed
OpenAI shut down its "Preparedness" team, which evaluated whether the company's own AI models could pose catastrophic risks.
Preparedness responsibilities were redistributed to existing groups and several safety staffers left.
The Decoder · attributed
The work has been parceled out to existing groups, and several safety staffers have left.
The dissolution triggered internal unrest and staff departures at OpenAI.
The Decoder · attributed
This move has triggered internal unrest and staff departures
Thrive Capital's 2022 early-stage $516M fund reached over $3.7B net value by end of June.
Techmeme · attributed
Investor letter: Thrive's 2022 early-stage $516M fund that invested in OpenAI and SpaceX was worth $3.7B+ as of the end of June, net of fees and other charges
Thrive's 2022 early-stage fund achieved a 7.2x net multiple in under two years driven by OpenAI and SpaceX positions.
Techmeme · attributed
Thrive's 2022 early-stage fund achieved a 7.2x net multiple in under two years, validating the extreme concentration of value in a handful of AI and space assets.
Why it matters
Teams shipping on OpenAI models cannot rely solely on the vendor's internal safety structure; independent audits and external oversight become more important when dedicated catastrophic-risk staff depart.
Limits and uncertainties
The packet does not identify which existing OpenAI groups now own Preparedness responsibilities.
Several source excerpts in the packet are truncated and do not provide full article text.
Thrive fund performance figures are attributed via Techmeme to a Bloomberg investor letter, not primary OpenAI disclosure.
Practical implications
Operators should not assume a centralized catastrophic-risk review function remains at OpenAI after Preparedness was dissolved.
Builders may need independent third-party audits and stricter external regulatory frameworks when vendor internal safety assurances become less visible.
What to watch
Further departures from OpenAI safety staff following the Preparedness dissolution.
Public identification of which internal groups now conduct catastrophic-risk evaluation at OpenAI.
Whether OpenAI publishes updated catastrophic-risk review criteria or governance after redistribution.
