To keep the confidentiality, integrity and availability of AI systems and services and assure safe and secure use at all times, take measures that are reasonable against the technical standard of the time; understand the characteristics of the systems and examine whether the connections between systems needed for normal operation are properly made; and recognise that vulnerabilities cannot be removed completely, given that subtle perturbations mixed into inference data can lead to judgments the relevant stakeholders did not intend.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.