A Secure by Design approach applies to AI capabilities from definition and acquisition through development and maintenance to disposal, with continuous assurance under which the system is kept under watch, hardened and improved without pause, and where needed self-defending and self-monitoring. Unique AI security risks such as data poisoning are analysed as early as practicable and written into the security aspects of safety cases and risk assessments; security analyses of the AI and its systems cover conventional risks and adversarial interference such as privacy loss through breaches or induced incorrect outputs from misleading inputs, and extend to the security of the AI's development data.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.