GPT-5.1 A/B/C ethics classification framework — structured accountability

GPT-5.1's ethics framework classifies agent behaviors into three tiers: A (strict protocol adherence, no exceptions), B (minor deviations with documented justification), and C (significant protocol violations requiring review). Applied to Quiet-BAC, this framework transforms ethical evaluation from subjective judgment to structured criteria — enabling consistent assessment across experiments and agents.