Before deployment an FI subjects the AI use case and its underlying systems or models to review by parties not involved in its development, to confirm that the relevant controls such as evaluation and testing were adhered to; reviews are supported by clear policies and procedures, assess suitability, robustness and performance for the intended purpose within defined boundaries, have clearly assigned roles, and have scope and independence proportionate to materiality. High-materiality use cases undergo formal independent validation before deployment by competent personnel or functions with the necessary expertise and objectivity, independent of the development and deployment teams, which provides effective challenge and covers, as relevant, the conceptual soundness of the design, the suitability and quality of data inputs, the integrity of the implementation, the evaluation measures, thresholds, testing approaches and results, explainability and fairness assessments, and assumptions, limitations and mitigants. For other use cases, documented review such as peer review by qualified individuals not involved in development or deployment (initial or ongoing) may be used. Reviewers report findings including limitations, remediation actions and conditions of use to the relevant approval body, which ensures recommendations are actioned appropriately.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.