Mechanisms and protocols at deployment measure the fairness and performance of outcomes and their impact on individuals and communities, checking whether represented groups receive benefits in similar proportion and whether the system disproportionately harms members of any demographic. The predefined fairness metrics are monitored in production, deviation beyond the threshold triggers investigation of whether the model must be renewed, and the overall harm and benefit to sensitive groups is quantified.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.