OpenAI unveils incident-tracking plan alongside six new AI safety examples
OpenAI disclosed six additional incidents of unexpected or concerning AI behavior and announced a framework to track, investigate, and publicly disclose misalignment, including a process for developers to flag issues for review and a preference for transparency even when the significance is uncertain.