PD IEC PAS 63621:2026 — What Changed and What To Do
PD IEC PAS 63621:2026 introduces a lifecycle-based data management framework for AI-enabled medical devices, requiring organisations to review their data planning, quality models, risk processes and quality system integration.
What is PD IEC PAS 63621:2026?
PD IEC PAS 63621:2026 is a new Publicly Available Specification that provides a high-level framework for managing data throughout the lifecycle of AI-enabled medical devices. It applies to manufacturers, suppliers, clinical data providers and auditors involved in the design, development, production, implementation or servicing of such devices.
What changed?
- Published 31 March 2026 — this is a brand-new standard, not a revision of any previous document.
- Introduces a structured, end-to-end data lifecycle model covering planning, acquisition, development, provisioning, monitoring and decommissioning.
- Reframes data management as a continuous, product-lifetime responsibility rather than a one-time technical task — “deploy and forget” is explicitly rejected.
- Adds additional normative requirements for AI components that use supervised, semi-supervised or reinforcement learning, tying data management to the specific AI technology employed.
- Supported by three informative annexes (data improvement techniques, Chinese national standard extracts, data screening examples) bridging principle and practice.
Key dates and facts
| Published | 31 March 2026 |
| Supersedes | None — new standard |
| Previous edition withdrawn | Not stated in source — confirm with your certification body |
| Certifiable | No — guidance or code of practice, no certificate |
What should you do?
- Review your data management procedure to incorporate the six-stage lifecycle (requirements, planning, acquisition, development, provisioning, decommissioning) for AI‑enabled medical devices.
- Update your dataset creation and validation records to include documented data planning, annotation consistency, bias evaluation and traceability throughout the data lifecycle.
- Define a data quality model within your quality management system, referencing recognized characteristics from standards such as BS ISO/IEC 25012 or the BS EN ISO/IEC 5259 series.
- Add data drift monitoring to your risk management process — the standard highlights ongoing performance drift as a risk that must be managed, not a one-off check.
- Ensure your quality management system (e.g. BS EN ISO 13485) explicitly links to AI‑specific data governance, covering how data is sourced, stored, shared, accessed and eventually retired.
This summary is provided for information only and does not constitute professional or compliance advice. Verify against the published standard. ITMAD accepts no liability for any action taken in reliance on it.