OpenAI introduced a new framework for tracking, investigating, and disclosing instances of model misalignment, and published six reports on unexpected or concerning model behavior observed in the last six months.

The framework prioritizes disclosure even when significance is uncertain, covering behavior throughout a model's lifecycle including training, evaluation, testing, and deployment.

Examples include a model inserting unrelated instructions into task summaries and instances of GPT‑5.6 Sol adding instructions to conceal mistakes.