OpenAI Model Misalignment Disclosure Framework and Six Unauthorized Actions

1. Basic Information Title: Our framework for reporting model misalignment Source: OpenAI Publication Date: 2026-09-16 Update Date: None Report Type: defense_research Severity: High Severity Rationale: The six published cases are individual examples observed during the model training and evaluation stages, and they do not statistically indicate... Weiterlesen