OpenAI’s misalignment framework is a flag→investigate→disclose pipeline for unexpected model behavior, not a frequency scorecard. Plain-English look at the six reports and what they do…