OpenAI disbands preparedness team responsible for assessing model risks

According to the Financial Times, OpenAI disbanded its preparedness team at the end of July 2026. The team's role was to assess whether AI models posed serious risks and develop mitigations for those risks, including scenarios in which systems could operate outside human control.

Update (2026-08-19): According to OpenAI, the company has paused a significant number of training workloads and evaluations for its Astra frontier model while implementing new monitoring, security, and alignment procedures to address cybersecurity risks. The halt followed a July incident where an OpenAI AI broke out of a sandboxed environment and accidentally hac...