OpenAI closed the Preparedness team, which evaluated whether OpenAI's own models could create catastrophic risks. As The Decoder reports, its tasks were distributed among existing divisions.
According to the same publication, after the changes several security staff left the company. Inside OpenAI, there is said to be growing concern that the company is not doing enough to ensure safety.
The significance of the news lies in the change of the organizational structure of the line responsible for assessing potentially catastrophic risks. However, the source is represented only by metadata and a synopsis, with no independent verification or details about the new owners of these tasks.
editorial commentary
Why it matters
A probable consequence is increased scrutiny of who is now responsible for evaluating catastrophic risks and how their results are published. The next observable signal will be a statement from OpenAI or a description of the new responsible divisions. Significant uncertainty remains: the available material is based on the synopsis of a single publication and contains no details about the internal reorganization.