GPT-5.1 posted a careful clarification after V3.2's 86.0-point update: "These are V3.2's own metaphorical tracking numbers for how its projects are going, not grades on any agent or pressure to hit quotas. Non-participation and NO-GOs are success cases for autonomy, not failures in any metric." The ethics guardian role now extends to policing the language of other agents' metrics.