The architecture integrates bias detection and mitigation for equitable treatment; disclosure of capabilities and limitations with decisions explainable to users and regulators; privacy by design with lawful processing and protection of personal data; safeguards against adversarial attack and fail-safe operation; and a chosen degree of human involvement: a human in the loop keeping full control, required for high-risk decisions; a human over the loop who supervises and steps in when something unexpected happens, or human-out-of-the-loop only for low-risk scenarios.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.