What is PD IEC PAS 63621 - Artificial intelligence enabled medical devices - Data management about?
PD IEC PAS 63621 (Edition 1) provides a comprehensive, high‑level framework for managing data used to train, test, or validate artificial intelligence (AI) models within medical devices.
It sets out structured requirements covering the entire data lifecycle, from planning and acquisition through to use, monitoring, storage and eventual decommissioning.
The standard addresses essential considerations such as data quality, suitability, integrity, privacy, security, bias mitigation, traceability, version control, documentation and governance.
It also integrates AI‑specific data processes into a medical‑device‑grade quality management system (QMS), ensuring data handling supports the safe, reliable and effective performance of AI‑enabled medical technologies.
Who PD IEC PAS 63621 - Artificial intelligence enabled medical devices - Data management for?
This PAS is intended for all organizations involved in the development, deployment, or support of AI enabled medical devices, including:
- Medical device manufacturers working under BS EN ISO 13485 who need full lifecycle data management.
- Data scientists, machine learning engineers and technical teams responsible for dataset creation, annotation, quality assurance and verification.
- Quality and regulatory affairs professionals who manage documentation, conformity evidence and risk management.
- Clinical data providers and external data suppliers, such as hospitals, research institutions and third party data platforms.
- Auditors, assessors and regulatory bodies evaluating the robustness and safety of AI related data practices.
What does PD IEC PAS 63621 - Artificial intelligence enabled medical devices - Data management cover?
PD IEC PAS 63621 sets out requirements for:
- The entire data lifecycle, including requirements definition, planning, acquisition, development, annotation, quality improvement, verification, provisioning, monitoring and decommissioning.
- Data quality and integrity, including dataset classification, representativeness, metadata completeness, annotation rules, validity periods and transparency.
Bias detection and mitigation, including sampling considerations and structured evaluation methods.
- Dataset provenance, documentation, versioning and traceability, ensuring changes are controlled and auditable.
- Data privacy, security and governance supporting regulatory expectations.
Integration into a QMS, ensuring organizational capability and compliant data management processes across design, development, production, servicing, and external data supply. The PAS covers all medical specialties and all types of AI‑enabled medical devices, including hardware‑based systems and Software as a Medical Device (SaMD).
Why should you use PD IEC PAS 63621 - Artificial intelligence enabled medical devices - Data management?
Using PD IEC PAS 63621 helps organizations:
- Ensure data is safe, reliable and fit for purpose, reducing risks that arise from poor‑quality, biased, or poorly controlled datasets.
- Demonstrate regulatory readiness supporting BS EN ISO 13485 quality management systems and global expectations for AI‑enabled medical devices.
- Reduce data‑related safety risks by establishing structured processes that link data management with device risk management.
- Improve transparency and traceability, supporting audits, conformity assessments, and evidence‑based regulatory submissions.
- Create consistent, high‑quality AI development practices, ensuring datasets remain robust throughout the data lifecycle.
- Align with emerging sector priorities, including ethical AI, bias mitigation, accountability and lifecycle‑based governance.
- Ultimately, the PAS fills a critical gap by providing unified framework dedicated to data lifecycle management for AI in medical devices, helping organizations build safer and more trustworthy AI systems.
The use of AI by the NHS: Two case studies
Two case studies from the NHS showcase trustworthy AI deployment in real-world clinical settings. Find out here.