UK Updates Regulatory Framework for Software & AI as Medical Devices: Key Changes Ahead for EBME Professionals

UK Updates Regulatory Framework for Software & AI as Medical Devices: Key Changes Ahead for EBME Professionals

In a significant step forward for medical technology regulation, the UK’s Medicines & Healthcare products Regulatory Agency (MHRA) has revised its guidance on Software and Artificial Intelligence (AI) as Medical Devices to ensure safer innovation and clearer oversight for devices that increasingly rely on software, machine learning, and AI.

The guidance, updated on 3 February 2025, is especially relevant to EBME (Electronics, Biomedical, and Medical Engineering) professionals involved in procurement, maintenance, and deployment of AI‐embedded medical devices, as it clarifies classification, post‐market vigilance, and regulatory responsibilities.

What’s New & What You Need to Know

1. Reinforced Role of the Software Group
MHRA’s Software Group is now more prominently tasked with ensuring that software as medical devices (SaMD), including AI as a medical device (AIaMD), meet clinical safety and performance needs. Among their duties are reviewing technical documentation, overseeing clinical investigations, and supporting both pre‐market and post‐market phases.

2. The “Change Programme” & Roadmap
Central to the update is the Software & AI as a Medical Device Change Programme. This roadmap outlines reforms across the full lifecycle of regulated software: from how devices are qualified and classified, through requirements for evidence, to transparency (interpretability, explainability) and how adaptivity (when AI models change or retrain) will be managed.

3. Classification & Qualification Guidance
There is revised guidance to help manufacturers and developers determine whether their software qualifies as a medical device (or an in vitro diagnostic) under UK law. Also included is new material for digital mental health technologies around how to characterise these devices, qualify them, and classify them appropriately.

4. Post‐Market & Vigilance Requirements
Guidance has been updated on reporting adverse events and managing safety in the field. EBME technicians and biomedical engineers need to be aware of obligations under the MHRA’s Yellow Card scheme (including app reporting) and how software behaviour (including AI changes) post‐deployment must be monitored.

5. AI Specifics: Adaptivity, Transparency & International Alignment

• The MHRA emphasizes the need to manage adaptive algorithms properly — when an AI model learns or changes after deployment.
• There is also direction toward transparency: ensuring that systems are explainable/ interpretable, and that change control plans are in place.
• The UK is aligning with international regulators (notably FDA, Health Canada) on principles for good machine learning practice, transparency, and control plans.

 Implications for EBME Expo Attendees, Suppliers, & Hospitals

• Device Maintainability & Updates: Biomedical engineers will need processes in place to track and validate software/AI updates in deployed devices, especially adaptive AI, to avoid regulatory non-compliance.
• Procurement Decisions: When purchasing or procuring devices, clarity on whether the device qualifies as AI/Software medical device—and if so, its classification—is now more essential.
• Safety Reporting: Duties for reporting incidents or malfunctions involving software or AI are more clearly defined; staff training will be needed so that it’s understood when and how to report via Yellow Card or local equivalents.
• Collaboration with Manufacturers: Expect more formal documentation, for example about intended use, change control plans, and explainability. Supplier guarantees and evidence will likely form part of tender requirements or service contracts.
• Quality Assurance & Testing: QA teams will need to incorporate evaluation of software behaviour under retraining / adaptive situations, and perhaps more rigorous pre‐market clinical evidence.

What’s Next & What Remains Under Development

While the revised guidance and the change programme provide a clearer trajectory, there are still areas under development:

• Defining “significant change” when AI models adapt dynamically, and how such changes trigger regulatory review. MHRA has been working with Brunel University on this.
• Deliverables in the roadmap: several guidance documents are still in draft or consultation form. EBME professionals should stay alert for updates.
• The digital mental health technology project is a three‐year programme funded by Wellcome Trust, so further specific guidance is expected.