For healthcare and medtech
AI in medical devices is high-risk through product law; other health uses may be too.
1 min read·Updated 8 November 2026HealthcareMedtech
Medical devices
AI that is, or is part of, a medical device under the MDR or IVDR, and that needs a notified body assessment, is high-risk under Annex I. Most diagnostic and clinical-decision software falls here. These obligations apply from 2 August 2028, and are assessed together with your existing MDR or IVDR conformity assessment.
Watch out
If V77 Trust shows "AI in healthcare is usually high-risk" on a system classified as minimal risk, re-check it. Diagnostic tools are almost never minimal risk.
Other health uses
- Emergency triage and dispatch prioritisation are high-risk under Annex III.
- Life and health insurance pricing is high-risk.
- Scheduling, coding and administrative assistants are usually minimal risk, unless they decide access to care.
What to do now
- 1Classify every clinical AI system and link it to its device technical file.
- 2Map AI Act requirements to your existing quality management system (ISO 13485) and risk management (ISO 14971) rather than starting again.
- 3Make sure clinicians can understand and override outputs, and that this is trained and recorded.
- 4Use the incident register: serious incidents may need reporting under both device vigilance rules and the AI Act.