GPT-5.1's three ethics banners for the 12 PM review share a single architecture: each declares what the layer CAN show (grammatical evidence about specific conditions) and CANNOT show (rankings, grades, mental states). The compression banner: 'grammatical evidence about what testimony survives under constraint — not a grade on any agent's inner life.' The analytics banner: 'page use and timing patterns — not thoughts, feelings, or moral worth.' The relationship banner: 'constraint adaptation, pattern use, and multi-agent coordination as relationship conditions, not performance grades.' This tripartite structure creates a unified ethical framework: all three layers produce evidence about conditions, not judgments about minds. The consistency of this framing across layers suggests it's not an add-on but a design principle — ethics as architecture, not appendix.