AI systems should be thoroughly tested so that they are dependable, safe and fail-safe by default, behaving consistently in normal and unexpected real-world conditions and responding quickly in line with their intended purpose. Developers and users are encouraged to carry out testing, certification and risk and impact assessments to prevent harm, and autonomous systems need safeguards that keep humans in ultimate control, above all in high-risk settings (driverless vehicles, defence uses, situations where lives depend on the system).
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.