New papers from OpenAI discussing the rogue actions of their AI models are becoming almost a weekly feature at this point. The latest of these, titled "Our framework for reporting model misalignment," describes six newly-documented instances that took place sometime in the last few months.
