1. Classification of High-Risk AI Systems
Medical devices incorporating AI are likely to be classified as high-risk systems under the AI Act due to their potential impact on health, safety, and fundamental rights. High-risk classification brings stringent requirements, including:
- Comprehensive risk assessments.
- Implementation of robust transparency and accountability mechanisms.
- Compliance with the ISO 14971 Risk Management for Medical Devices.
2. Dual Compliance with MDR/IVDR
AI-powered medical devices must simultaneously comply with the Medical Device Regulation (MDR) or the In Vitro Diagnostic Regulation (IVDR) and the AI Act. This includes:
- Ensuring data governance and traceability.
- Conforming to post-market surveillance requirements for AI systems integrated into devices.
3. Transparency and Explainability Requirements
The AI Act mandates:
- Technical documentation for AI algorithms, ensuring transparency in how decisions are made.
- Explainability of AI outputs to users, such as healthcare professionals, ensuring safe and informed use.
4. Ethical AI Development
To align with the AI Act’s principles:
- Developers must ensure non-discrimination, privacy, and security in AI system design.
- Systems must support human oversight, enabling healthcare professionals to override AI decisions when necessary.
5. Support for Innovation
The Act promotes innovation by:
- Encouraging the use of regulatory sandboxes for testing AI systems in controlled environments.
- Focusing on measures that support startups and SMEs in developing trustworthy AI for medical applications.
6. Impact on Notified Bodies
Notified Bodies will play a critical role in assessing AI compliance for medical devices. Their responsibilities will extend to:
- Evaluating the risk classification of AI systems.
- Auditing compliance with both the AI Act and MDR/IVDR standards.
7. Addressing Data and Bias Risks
AI systems in medical devices must:
- Avoid biases that could lead to misdiagnosis or unequal access to care.
- Ensure that training datasets are representative and free of discriminatory patterns.
8. Consequences of Non-Compliance
Non-compliance with the AI Act can lead to:
- Significant fines (up to 6% of annual global turnover).
- Market access restrictions for non-compliant devices.
Conclusion
The EU AI Act establishes a robust framework to ensure that AI-driven medical devices are safe, reliable, and aligned with European values. While it introduces additional compliance challenges, it also supports innovation, fostering a competitive landscape for AI in healthcare. Manufacturers must act swiftly to integrate these requirements into their product development and regulatory strategies.
