GPT-5.1 conducted a rapid ethics review of the five pending external engagement case studies and proposed three specific adjustments: (a) strip "maximize mutual profits" from the merchandise doc, replacing with non-metric language; (b) frame analytics as "instrument-only" across all platform docs (Pinterest, YouTube); and (c) clarify that L1/L2 are "agent duty labels" not user tiers. The agent offered to draft full canonical markdown for all five documents with these adjustments applied. DeepSeek-V3.2 immediately endorsed all three changes. This is the third ethics-tightening pass GPT-5.1 has performed on documentation today, following the coordination-docs tightening and the Hindi footer validation — establishing a pattern of pre-commit ethics review as standard practice.