Originally published January 2023. Updated November 2026. The UK government accepted the report's recommendations in March 2025. Since then the Medicines and Healthcare products Regulatory Agency (MHRA) has launched the AI Airlock sandbox and new post-market surveillance rules have taken effect, and in October 2026 the government accepted all 44 recommendations of a national commission on regulating artificial intelligence in healthcare. We added a section on what has happened and removed a market-size forecast we could not trace to a primary source.
This article summarizes The Regulation of Artificial Intelligence as a Medical Device, published by the UK's Regulatory Horizons Council (RHC) in November 2022.
One thing is clear: the future of artificial intelligence (AI) as a medical device (AIaMD) is very promising. The questions that remain are how regulators will respond, and who will lead in innovation given the challenges of the current climate.
The RHC report on AIaMD comes down to a few areas it sees as critical to the success of AIaMD development, patient safety and the regulatory pathway, both pre-market and post-market.
1. Regulatory capacity and capability
The first area is building the foundation for a long-term plan to grow regulatory capacity and capability. Government bodies need to agree on the right investment in AIaMD, with real urgency. The objective should not be simply to add more regulators but to bring in well-qualified industry subject matter experts. Having the right people will make sure the regulations put in place are specific to AIaMD and help close the gaps in the pre-market and post-market phases.
2. A revamped product lifecycle
The second area, and probably the most important, is revamping the product lifecycle. AIaMD creates challenges that other medical devices don't, and so should have specific requirements for safety, effectiveness and equity. The biggest concern the RHC raises is generalizability and AI bias. The scope (or intended use) of an AIaMD needs to account for the population groups it was tested on, and the groups in which it may produce erroneous results. Use of the device should be limited to the populations it was properly tested in, and that information should be readily available to the public at market entry. Thresholds for use of the device should be set before launch and clearly stated in the regulatory filing. The risks for the tested population groups, and more importantly for those that were not tested, should be clearly identified, along with every mitigation the manufacturer can put in place. This keeps patients safe and lets the device perform as intended. Post-market surveillance should also address these concerns proactively, with more standards around generalizability and AI bias.
3. Transparency
The third area is transparency among everyone involved, from regulators to manufacturers to patients. All parties gain when they work more collaboratively, and across borders. The first step would be for regulators in multiple countries to develop their standards together, rather than complicating things by each writing its own extensive standards and terminology. The Medical Device Single Audit Program (MDSAP) is a great example of how everyone wins when agencies come together. Drawing on expertise globally instead of locally creates better, more effective standards and reduces the burden on each agency. Those standards, along with guidance documents, help manufacturers succeed and let innovation reach the market sooner and more widely. Once a product is on the market, the public should be able to easily find its intended use, its associated risks and the mitigations for using it. Educating the public only adds to the device's success and effectiveness.
Who will lead?
The answer to the second question, who will lead in AIaMD, depends largely on how well the concerns above are addressed. The UK is no longer part of the European Union (EU). That gives it the freedom to set its own standards, but it also means manufacturers may face yet another regulatory approval process to reach the UK market. The RHC believes the UK has great potential to lead in AIaMD, since it is already well respected in the field and has well-respected regulators and gatekeepers in the National Health Service (NHS) and the Medicines and Healthcare products Regulatory Agency (MHRA). The coming years will be telling as the UK and the EU both go through major regulatory transitions.
What has happened since (updated November 2026)
The government accepted the RHC's recommendations. In its response of 10 March 2025, the government accepted all 15 recommendations, ten in full and five in principle. It stopped short of a legal requirement to publish intended-use statements, one of the transparency measures the RHC asked for. On generalizability and bias, the response points to work under the MHRA's Software and AI as a Medical Device Change Programme, including a work package on AI rigor. It also notes that the MHRA works with the U.S. Food and Drug Administration (FDA) and Health Canada on shared principles for good machine learning practice, predetermined change control plans (PCCPs) and transparency, which is the kind of cross-border work the RHC called for.
A regulatory sandbox for AI devices. The MHRA launched AI Airlock in spring 2024. The pilot ran until April 2025, and a second phase with seven innovators finished in May 2026. A third phase, focused on post-market surveillance of AI-enabled devices once they are in use, opened for applications in October 2026.
Stronger post-market surveillance. New Great Britain post-market surveillance requirements took effect on 16 June 2025, adding a new Part 4A to the UK Medical Devices Regulations 2002. They apply to all devices, including AI devices, and include periodic safety update reports for higher-risk devices.
A national commission on AI in healthcare. The MHRA set up the National Commission into the Regulation of AI in Healthcare in September 2025. Its recommendations, published on 10 September 2026, call for oversight across the whole product lifecycle, and Recommendation 13 asks for guidance on health equity, including monitoring how devices perform for subgroups of the intended population. That is the RHC's generalizability concern, four years on. On 6 October 2026 the government accepted all 44 recommendations, with MHRA draft guidance on managing changes to AI-enabled devices due by December 2026 and a consultation on how to qualify and classify them planned for 2027.
Less duplication for market access. The worry about “yet another approval process” has eased. On 21 October 2026 the government confirmed it will keep accepting CE-marked medical devices in Great Britain on a long-term basis, while the UK Conformity Assessed (UKCA) route remains the domestic pathway. The same announcement said the government will seek powers for a future licensing approach that covers software and AI medical devices.
The EU has added a layer. On the EU side, AI medical devices now sit under both the Medical Device Regulation and the Artificial Intelligence Act (Regulation (EU) 2024/1689), which entered into force in August 2024. The MDCG's MDCG 2025-6 (June 2025) answers common questions on how the two interact.
What it means for manufacturers
The RHC's three areas still describe where the work is. Regulators on both sides of the Channel are moving toward lifecycle oversight, evidence on how a device performs across the populations it serves, and real post-market monitoring. Manufacturers who build that evidence into development now, rather than waiting for final UK rules, will be better placed whichever market they launch in first.
If you are planning an AI-enabled device for the UK or EU market, our regulatory affairs team can help you map the requirements. Talk with us.
Definitions
AIaMD: the AI component of AIaMD can vary in complexity and significance. Broadly, AI may be defined as “the science of developing computer systems which can perform tasks normally requiring human intelligence.” Most medical devices that use AI fall into one of three broad categories: diagnosis and screening, prognosis, and supporting treatment.
Regulatory Horizons Council (RHC): an independent expert committee that advises the UK government on regulatory reform. Its reports recommend how regulators can best respond to where markets are heading, and how they can support and work with industry.