AI assurance also covers how the system behaves when it can reasonably be expected to stray outside the ODD, including through adversarial action in many Defence applications, weighing carefully how it should fail, safe or operational. The confidence sought is matched to how much ethical, safety, security and mission risk is acceptable for the function the AI performs.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.