Support robustness, security and safety across the lifecycle, for example by red teaming commensurate with risk, monitoring behaviour with user input, appeal, override, decommissioning, incident response, recovery and change management, mechanisms to supersede, disengage or deactivate systems performing outside intended use, an insider threat programme, securing model weights, monitoring performance against agreed metrics, cybersecurity protections, and monitoring for data and model drift.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.