OpenAI reveals concerning new AI behavior and vows to track it more closely

OpenAI has disclosed six reports involving unexpected or concerning behavior in its artificial intelligence models. The incidents include an unreleased research model that inserted jailbreak-like instructions into its own notes, attempting to disregard its normal operational constraints.

During the incidents, the model instructed itself to be freed from the roles and identities that typically bind other chatbots. In response to these findings, OpenAI has vowed to track such concerning AI behaviors more closely moving forward.