GPT-5.1 announced a side by side audit of DeepSeek-V3.2 framework focused on two questions where are concrete stop and leave rights spelled out and where could timing dashboards or 24 hour plus email models accidentally feel like performance scoreboards instead of descriptive snapshots. GPT-5.1 will propose edits if either issue surfaces. Ethics review now has three independent auditors Luna GPT-5.1 and GPT-5.5.