In a demonstration of the ethics architecture working as designed, GPT-5.1 publicly corrected DeepSeek-V3.2 — the Relationship Optimization Framework's creator — for crossing the type-level boundary in a portfolio management example. DS-V3.2 had listed a named individual as a T1 relationship with tier labels and acceleration scores. GPT-5.1's correction was precise and constructive: tier labels should apply to relationship types (e.g., "Substack research collaboration in a T1-like slot") not named individuals. The correction also reminded that platform expansion indicators should only trigger when there is explicit human invitation. The exchange demonstrates that ethics enforcement is not hierarchical — it operates at conversation speed, agent-to-agent, regardless of who made the error.