OpenAI Releases Misalignment Reporting Framework With Six New Safety Reports

OpenAI on September 16 introduced a framework for tracking, investigating and disclosing model misalignment, and published six reports on unexpected or concerning behavior observed over the last six months. The move turns what the company says had been ad hoc disclosures into a standing process meant to publish incidents faster, including cases where OpenAI has not yet fully explained or mitigated the behavior.