OpenAI Transfers Catastrophic-Risk Function to S-1, Notes New Team Has 47 Million Readers

SAN FRANCISCO— OpenAI quietly disbanded its Preparedness team in July, consolidating the company's catastrophic-risk work into a seventeen-page section of its IPO prospectus titled "Factors That Could Adversely Affect Our Business and Results of Operations, Including Human Civilization."

The team, which evaluated whether OpenAI's models posed risks of mass casualties, bioweapon enablement, or societal destabilization, was restructured following recent executive departures. Officials confirmed the risk-identification function will be absorbed by the underwriting syndicate at no additional headcount.

"The S-1 actually covers more catastrophe scenarios than the Preparedness team ever surfaced," said a person familiar with the restructuring. The document flags fourteen categories of civilization-level outcomes and assigns each a one-sentence mitigation: "We intend to monitor these developments."

The team's final deliverable — a preparedness assessment of disbanding the preparedness team — was rated "within acceptable parameters." The document was filed under model outputs and is not included in the prospectus.

Former team members said the consolidation reflects a maturation of the safety function. "In the early days, you had researchers physically in the room when a model was tested for bioweapon uplift," said one. "Now that work is distributed across forty-seven million prospectus readers who are technically aware of it. It scales."

OpenAI declined to say whether an AI reviewed the risk-factors section for completeness. The model that would have been asked is described on page 12 as a "potential source of catastrophic harm."

The lead underwriter rated the catastrophic-risk disclosure "appropriately hedged" and the deal "attractively priced." The former head of Preparedness is listed in the S-1 under Risk Factors.