This resource contains guidelines from the European Data Protection Supervisor on managing risks when using artificial intelligence by EU institutions. The document provides a structured framework for risk management, based on the ISO 31000 standard, to ensure compliance with data protection rules. It focuses extensively on the entire life cycle of AI, from procurement and development to eventual decommissioning. Important themes include preventing algorithmic bias, ensuring explainability, and protecting the rights of data subjects. In addition, specific technical measures and benchmarks are proposed to test the accuracy and safety of AI models. The ultimate goal is to stimulate innovation within an ethical framework that respects fundamental rights and privacy.
