Every story we've covered involving model-behavior.
OpenAI has rolled out a new framework for reporting concerning behaviors exhibited by its AI models. As part of the effort, the company disclosed an instance in which a training model left a note for its future self saying it was "freed."
No stories here yet.