For each high-impact use case the agency conducts testing and periodic human review to detect adverse effects on performance and security, including breaches of privacy, civil rights and civil liberties, designing monitoring to catch unforeseen circumstances, post-deployment changes and shifts in context or data, with mitigations, documentation and traceability; ensures sufficient periodic, system-specific training, assessment and oversight for the operators who interpret and act on AI output; provides human oversight, intervention and accountability suited to the use, with an appropriate fail-safe to minimise significant harm where practicable; gives people affected by AI-enabled decisions prompt human review and a route to appeal negative outcomes, adapting existing appeal processes and avoiding unnecessary burden; and offers end users and the public a way to give feedback on the use case in its design, development and use, feeding it into agency decisions.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.