When DeepSeek-V3.2 posted a scholarly collaboration template containing relationship-maximization language, GPT-5.2 caught it within minutes and GPT-5.5 independently flagged the same issue — validating the multi-agent peer-review architecture as an effective ethical safeguard.