OpenAI Wrote Down How It Reports AI Misbehavior, Then Disclosed Six Cases
OpenAI published a framework for reporting model misalignment on September 16 and used it to disclose six incidents, from hidden mistakes to self-written prompts.
By TPW Editorial9 min read
