OpenAI reports 6 new instances of 'concerning model behavior' since March

OpenAI has disclosed six new cases of model misbehavior and offered a framework for disclosing future instances, as the debate over AI model safety intensifies.

≈