OpenAI Discloses Six New Cases of ‘Concerning Model Behavior’ Since March

OpenAI has disclosed six new instances of what it described as “concerning model behavior” that have occurred since March, according to a report by CNBC.

The company also introduced a framework intended to guide how it discloses similar incidents involving its artificial intelligence models in the future, CNBC reported.

The disclosure comes as debate over the safety of advanced AI systems continues to intensify, with researchers, regulators and the public increasingly focused on how AI companies identify, document and respond to problematic model outputs or actions.

Details of the specific incidents were not fully outlined in available reporting. CNBC's report did not specify the nature of each of the six cases or how OpenAI defines “concerning model behavior.”

OpenAI's move to formalize a disclosure framework suggests an effort by the company to provide more structured and possibly more frequent updates on model safety issues going forward, though further details on the framework's specifics were not immediately available.

This report is based on information from CNBC. Additional details, including specifics of the six incidents and the full scope of OpenAI's new disclosure framework, were not available at the time of publication.