The MHRA has published findings from the National Commission into the Regulation of AI in Healthcare research, pointing to the need to balance a desire to use AI in improving patient care and supporting healthcare professionals, with “safe, fast and trusted” regulation.
A call for evidence ran from December 2025 to February 2026, receiving a total of 761 responses. This posed questions around required changes to the current regulatory framework, monitoring of AI products in healthcare, liability and responsibility.
Almost three-quarters of respondents (73 percent) disagreed or strongly disagreed that the current regulatory framework is sufficient to ensure safety and performance standards; with 61 percent disagreed or strongly disagreed with its effectiveness in tackling data governance and data privacy. 61 percent also thought current requirements for clinical evidence are insufficient, and 65 percent pointed to a need for more to be done on post-market surveillance.
Opinions on the impact on innovation were relatively split, with 65 percent of industry commentators considering it to be either “somewhat” restrictive or too restrictive; compared with 77 percent of the public who felt it was too loose or “somewhat” loose.
Current regulation is “not well suited to iterative and adaptive AI systems”, the Commission found, with stakeholders calling for a proportionate approach taking into account risk, patient safety, and fairness, as well as strengthened clinical evidence requirements and post-market surveillance. There is strong consensus about the need for regulatory reform, it goes on, with 34 percent of respondents calling for “significant reform”, and 35 percent calling for a “complete overhaul”. A healthcare professional is quoted: “The current UK pathway for getting AI and digital health technologies into the NHS feels too restrictive and creates unnecessary friction for safe, evidence-based innovation.”
Better post-market surveillance and monitoring is needed, respondents identified, indicating the shortcomings of a one-off approval process in measuring performance and risk, and the value of ongoing oversight across the lifecycle, taking into account performance drift, validation in real-world settings, and changes in performance over time. One response from an unidentified public body reads: “Post-market surveillance for AI as a Medical Device should include continuous false negatives and near-misses, to identify trends and emerging risks.”
A model promoting shared accountability was favoured by respondents over individual responsibility, the call establishes. One suggestion from an unnamed public body is that responsibility should be divided as follows: “Manufacturers designing AI systems that are transparent, usable, and robust in real clinical contexts; providers ensuring safe deployment and governance; and clinicians retaining accountability for clinical judgement.” Clearer and more consistent frameworks are needed to help address uncertainty and support safe use, it adds. “While existing legal frameworks may be able to resolve straightforward cases of harm, they create uncertainty in more complex scenarios involving AI and may fail to incentivise appropriate risk management for the full range of harms these systems can cause,” one response notes.
Concerns were expressed about how consent is currently managed for the use of health data, including by commercial entities, with governance, compliance, and fragmented data also seen as barriers to development and deployment. One response stated: “AI systems must meet strict standards for anonymisation, encryption, and public accountability. NHS patient data should not be used by private organisations without clear consent and demonstrable public benefit.”
Elsewhere, respondents highlighted the need for human oversight and responsibility for clinical judgement, transparency and explainability, and ongoing training around the understanding of AI use in healthcare. Incident reporting and learning mechanisms need to be improved, with “clearer communication” when things go wrong, the Commission notes, with challenges voiced about underreporting, lack of awareness of existing mechanisms like the Yellow Card scheme, and inconsistency in reporting across settings.
“Through the Call for Evidence, trust emerged as a core enabler of AI adoption in healthcare,” according to the Commission. “Patients and the public called for consistent involvement, consent, and clarity over the role of AI systems, whilst professionals highlighted the need to take a proportionate approach to explaining how AI is being used to patients. Providers advised that clear and consistent transparency and communication frameworks are needed whilst industry respondents recognised that trust is key for the uptake of AI systems in healthcare.”
Wider trend: Health AI
The National Institute for Health and Care Research has awarded £8 million to six innovations using AI and digital to speed up diagnosis and improve patient care. Innovations granted a share of the funding include SAMURAI-CT, an AI tool designed to detect “serious findings” from head CT scans with aims of reducing discharge times by more than 20 percent. It is currently being tested across Oxford University Hospitals, Royal Berkshire, University Hospitals of Derby and Burton, and NHS Greater Glasgow and Clyde.
The Medicines and Healthcare products Regulatory Agency has announced plans to launch a regulatory AI sandbox with the aim to explore how artificial intelligence can support medicines development and safety. The sandbox, funded by the Regulatory Innovation Office, will enable AI tools to be tested that have the potential to predict how medicines will be absorbed, processed, or whether they may cause harm. The sandbox will also be used to inform the MHRA about the reliability of whether AI tools can support in deciding the safety of new medicines.
The Government of Canada has published its national AI strategy: “AI for All”, promising free AI literacy training for all citizens, up to 250,000 new jobs by 2031, an AI Missions Programme with $200 million for projects improving health outcomes, and $700 million for access to public compute for SMEs via a new fund.
