Browsing: Model Misalignment Reporting Framework

OpenAI published a framework for tracking, investigating, and disclosing instances of model misalignment on September 16, 2026, alongside six reports on unexpected or concerning behavior the company said it observed during the training or evaluation of its models.