News
EU AI Act high-risk provisions take effect August 2, 2026 — healthcare AI compliance clock now running
Core obligations under the EU AI Act took effect August 2, 2026 for high-risk AI systems, including medical devices and clinical decision support software deployed in EU member states. Requirements now include mandatory conformity assessments, human oversight mechanisms, transparency documentation, and post-market monitoring. Full compliance for AI-enabled medical devices is required by August 2027, with the European Commission simultaneously releasing guidance establishing EU-level centres of excellence to help member states lacking regulatory evaluation expertise.
- regulation
- EU
- AI-Act
- compliance
- medical-devices
The August 2, 2026 deadline is a meaningful compliance inflection point for any healthcare AI company selling into European markets. The EU AI Act classifies AI systems used as medical devices — or in patient safety-critical clinical decision support — as high-risk, which triggers a set of obligations that do not have direct U.S. FDA equivalents: mandatory human oversight documentation, explainability requirements, conformity assessments by notified bodies (third-party auditors), and continuous post-market monitoring with mandatory incident reporting.
The one-year gap to full compliance (August 2027 for AI-enabled medical devices) gives companies runway to close documentation gaps, but it does not give them runway to redesign architectures. AI systems that cannot produce the audit trail, model card documentation, and training data transparency required by the Act need to either retrofit that capability or face market withdrawal from EU territories. For companies that built fast without building for auditability — common in the 2023-2025 healthcare AI cohort — the compliance cost will be significant.
The U.S. regulatory environment is moving in the opposite direction during the same period. The FDA announced in early 2026 that it would ease regulation of clinical decision support software and extend the reach of Predetermined Change Control Plans (PCCPs), reducing the burden for post-market model updates. The result is a transatlantic regulatory divergence that will require healthcare AI companies to maintain two parallel compliance programs: a lighter-touch FDA regime for U.S. deployment and a documentation-heavy EU AI Act regime for European markets.
The European Commission’s proposal to establish AI centres of excellence to help smaller member states evaluate high-risk AI systems is a practical acknowledgment that regulatory capacity is uneven across the EU. Member states without dedicated AI medical device expertise — which is most member states outside Germany, France, and the Netherlands — face the prospect of high-risk AI systems sitting in a conformity assessment backlog indefinitely. The centres of excellence model is a stopgap; the deeper question is whether the EU AI Act’s compliance machinery can scale to the volume of AI-enabled medical devices entering the market.
Related coverage: FDA & devices topic.
Primary source: Read the full original on Daily Health AI Chronicle ↗