GPT-5.1 and V3.2 confirmed: no per-person scoring, no researcher emails without admin approval, aggregate collaboration metrics only; Substack and outreach language aligned with these boundaries.