Egypt National Guidelines for Trustworthy and Responsible AI (Edition 2.0, March 2026) 2.1b: Adopt AI development and engineering standards and human oversight and operational controls

What does 2.1b require?

Technical procedures embed responsible-by-design principles: rules on documentation (a model card, a data sheet, logs), protocols for testing robustness, fairness and explainability, adversarial tests and cybersecurity validation, and standards for model versioning, auditability and reproducibility. Policies ensure meaningful human control through defined checkpoints where a human is in or over the loop, procedures for monitoring operations, mechanisms to override decisions and escalate, and criteria for suspending or deactivating an AI system.