The rapid evolution of Artificial Intelligence is transforming healthcare, presenting both significant opportunities and considerable challenges. The European Union has adopted a risk-based regulatory approach with the AI Act (Regulation (EU) 2024/1689) to balance innovation with fundamental rights protection. This law classifies AI systems into four risk levels where high-risk systems, such as those used in medical diagnostics and robotic surgery, are subject to stringent compliance and ex-ante assessments. Despite AI promises improvements in several areas of Medicine (e. g. drug discovery, medical imaging, early diagnosis, disease prevention) potentially enhancing efficiency and reducing medical errors, current critical concerns include automation bias, the lack of long-term prediction accuracy, potential inequalities in technology access, and algorithmic opacity. For these reasons, AI systems can also exacerbate existing forms of human vulnerability and introduce new ones through the interaction between inferences, predictions, and AI-generated content. Competent authorities play a crucial role in ensuring that AI systems for health are governed and implemented effectively, and are prepared with adequate resources to meet market demand for testing and evaluation. Controlled environments (i.e. regulatory sandboxes) facilitate the testing and verification of AI systems through collaboration among developers, healthcare providers, regulators, and patients. These ensure that new technologies are safe, effective, and ethical before widespread clinical adoption, mitigating risks and protecting human vulnerability. The role of Competent Authorities is crucial to guarantee the protection of fundamental rights, health, and safety.