OpenAI disbanded its Preparedness team, responsible for evaluating whether its artificial intelligence models could create severe or catastrophic risks. According to reports, the tech giant dissolved the team last month as part of a wider internal restructuring.
The team played a key role in identifying emerging risks from increasingly capable AI systems. Its work covered areas such as cybersecurity, biological threats and potential loss-of-control scenarios.
OpenAI's own framework describes Preparedness as a process for tracking advanced AI capabilities and preparing safeguards against severe harm. The shutdown does not imply that OpenAI has abandoned its AI safety framework. Experts suggest the company will transfer duties performed by the Preparedness team into specialized teams.
These teams will be involved in sectors like cybersecurity and biosecurity. Furthermore, the company will continue to retain the Frontier Governance Framework that focuses on cyber risks, CBRN, manipulation, and loss of control.
OpenAI continues to release safety assessment reports of its models. The GPT-5.6 report indicates that some of its models have achieved the capability level concerning cybersecurity and biological and chemical risks.
The move comes as AI systems become more capable of performing complex tasks and interacting with external systems. OpenAI has acknowledged that such capabilities can introduce new categories of risk.
The company says its governance and safety frameworks will continue to evolve alongside model capabilities. OpenAI is in the process of simplifying its organizational structure and focusing on its key products and business. The absence of a preparedness unit will not make the discussions about AI regulation and safety fade away.
The key question now is whether distributing Preparedness responsibilities across specialized teams can preserve the same level of oversight that a dedicated risk-assessment group provided.