Artificial intelligence (AI) is moving from pilots to practice across imaging, patient monitoring and clinical decision support. To bring these tools safely into everyday care, clinicians and developers need to understand how AI-enabled medical devices are regulated in Great Britain, what evidence is required, and how products are monitored after launch. This article explains the current UK framework and the key changes introduced in 2025.
The UK framework at a glance
All medical devices in Great Britain are regulated by the Medicines and Healthcare products Regulatory Agency (MHRA) under the UK Medical Devices Regulations 2002, as amended after Brexit.
Software, including AI, is considered a medical device if it is intended for diagnosis, monitoring or treatment, and must meet the same safety and performance requirements as any other product.
The MHRA provides detailed guidance on Software and Artificial Intelligence (AI) as a Medical Device, setting out expectations for clinical evidence, usability, data management and control of algorithm updates.
CE and UKCA marking
Great Britain currently accepts both CE and UKCA markings. In 2025, the government confirmed its intention to allow indefinite recognition of CE-marked devices, ensuring continuity of supply while the UK’s own regulatory system evolves.
The MHRA is also expanding international recognition and reliance routes, which make it easier for products approved by trusted regulators, such as those in the EU or US, to reach UK patients more quickly.
What changed in 2025: stronger post-market surveillance
From 16 June 2025, new Post-Market Surveillance (PMS) requirements came into force in Great Britain. Manufacturers of devices placed on the market after this date must:
- run robust PMS systems that actively track device performance
- submit defined PMS reports to the MHRA
- identify and trend incidents
- take timely corrective and preventive action when issues are detected
For AI-enabled devices, this matters because models often evolve through software updates. The MHRA now expects manufacturers to closely monitor real-world performance, maintain transparent change-control processes, and report promptly if updates could affect safety or outcomes.
Classifying AI devices
AI technology itself doesn’t determine a device’s class – risk does. Tools that diagnose conditions, guide treatment or influence clinical decisions are usually Class IIa or higher. These higher-risk products require independent assessment by a UK Approved Body and must provide more extensive clinical evidence.
Evidence and documentation
Whether a device carries a CE or UKCA mark, it must show it is safe, effective and usable in real-world settings. Manufacturers should provide:
- Representative data proving reliable algorithm performance across diverse UK populations
- Clinical evaluation demonstrating measurable benefit and managed risk
- Usability testing so clinicians understand outputs and limitations
- Cybersecurity and data-governance safeguards
- Change-management plans explaining how AI updates are controlled, validated and communicated
The MHRA’s Software and AI Change Programme outlines how evidence requirements are evolving for adaptive algorithms.
Step-by-step: bringing an AI device to the UK market
- Confirm it’s a medical device
Check the MHRA’s Software as a Medical Device guidance to see if your product is regulated. - Determine its risk class
Classification defines the level of evidence needed and whether an Approved Body must review it. - Prepare technical documentation
Include clinical data, usability studies, cybersecurity measures and AI-specific change-control procedures. - Secure CE or UKCA marking
Both are currently valid. CE recognition is expected to continue indefinitely once consultation concludes. - Register with the MHRA
All manufacturers must list themselves and their products on the MHRA registration system. - Plan for PMS from day one
Set up monitoring and reporting aligned with the June 2025 PMS rules.
What’s next for UK AI regulation
The MHRA is developing a refreshed, AI-ready regulatory framework that will:
- clarify expectations for adaptive and learning algorithms
- strengthen transparency and post-market performance monitoring
- widen international reliance routes
- embed indefinite CE recognition (subject to consultation)
These reforms aim to balance innovation with patient safety and public confidence.
Why this matters for clinicians
Understanding regulation helps clinicians:
- choose devices with proven safety and performance
- plan safe pilots and procurement within NHS trusts
- ask informed questions about data, validation and software updates
- report performance issues effectively under the new PMS regime
Being confident in the regulatory pathway means faster, safer adoption of AI tools in everyday care.
Summary
AI-enabled medical devices in Great Britain are regulated under the UK Medical Devices Regulations 2002, with strengthened post-market surveillance from June 2025. Both CE and UKCA marks remain valid, and indefinite CE recognition is proposed.
Manufacturers must prove safety, transparency and control of AI updates, while clinicians should check compliance and ensure robust evidence before adoption. Understanding these pathways supports confident, compliant and patient-centred use of AI in UK healthcare.
