For each AI system, create and document monitoring requirements, including human oversight, before deployment, and evaluate them as part of continual improvement.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.