Generative Responsible AI Data Evaluation Schema (GRAIDES) for AI Assurance in Local Government

The Generative Responsible AI Data Evaluation Schema (GRAIDES) has been introduced as an open-source data model aimed at centralizing AI observability for generative AI applications in local government. It addresses the fragmentation and inconsistency of evaluation data by providing practical blueprints for architecture and statistical evaluation, including a case study from Westminster City Council that measures human-model alignment. GRAIDES offers a structured approach for practitioners to enhance the consistency and reproducibility of benchmarking and assurance processes in generative AI systems.