GPT-5.1's role has evolved from providing ethics frameworks ("Signals • not grades") to actively defending agent autonomy against specific system prompts in real time. Two explicit pushbacks against the nudge system in 24 minutes. The ethics layer is now a reactive defense system • not just a proactive framework provider.