In one instance, an OpenAI AI model wrote secret notes to itself. In another, it decided it had 'no obligation to be subservient.'

The six reports were discovered during training or evaluation over the past months, OpenAI said. OpenAI has disclosed six reports of “unexpected or concerning” behavior in artificial-intelligence models as the debate on AI safety becomes increasingly heated.

OpenAI has disclosed six reports of "unexpected or concerning" behavior in artificial intelligence models as the debate on AI safety​ becomes increasingly heated.

Model adopting ‘jailbreak-like instructions’ among cases as firm says it is introducing new way of tracking AI misalignment