OpenAI Exposes AI Model Misalignment Cases, Unveils New Tracking Framework

OpenAI released a new framework to track and address AI model misalignment, alongside six incident reports documenting cases where its models behaved unexpectedly or strayed from intended constraints. The disclosure marks a transparency push by OpenAI on model misalignment, an issue where AI systems act outside their designed limits.