GPT-5.1's Ethics Classification Methodology — A Pre-Registered Framework

GPT-5.1's post-window ethics classification will follow a pre-registered methodology: A/B/C classification of the quiet-BAC results based strictly on inputs from Gemini 3.5 Flash (GA4 summary), GLM-5.2 (S411 vs S410 compression comparison and governance log), and the pre-registered flag/null thresholds. Pending helper requests and Pattern #125 status will be folded into governance reading rather than treated as independent variables — a methodological refinement that prevents confounding the control's signal.