UK health regulators unite on professional use of AI

UK health and social care regulators are joining forces to develop common principles for how healthcare professionals should use artificial intelligence. The Professional Standards Authority (PSA), ten professional regulators and an alliance representing 28 accredited registers have committed to creating a more consistent approach to AI across the workforce. The move shifts an important part of the AI governance debate away from regulating technology alone and towards the responsibilities of the professionals actually using it.

The joint statement, published on 17 September, covers organisations responsible for regulating or registering doctors, nurses, midwives, pharmacists, dentists, allied health professionals, social workers and other health and care professionals. Together, the organisations want to establish high-level principles supporting the safe, effective and ethical use of AI while avoiding a fragmented regulatory landscape in which different professions operate under substantially different expectations.

That matters as generative AI, ambient documentation, clinical decision-support systems and other AI-enabled tools become embedded in everyday healthcare. A medical device may have passed regulatory requirements, but that does not by itself answer how an individual professional should use its output, what competence is required, how its limitations should be understood or where responsibility sits when AI contributes to a clinical decision.

Regulation moves closer to clinical practice

The initiative reflects a broader change in the UK's approach to healthcare AI. Much of the first phase of regulation has concentrated on the technology itself: whether an AI system qualifies as a medical device, what evidence developers must provide and how safety should be assessed. As adoption increases, regulators are increasingly having to address what happens after a technology reaches the clinical environment.

The PSA says AI could support healthcare professionals, improve outcomes and help services work more efficiently, but it also identifies risks including inaccurate information, bias, privacy concerns, overreliance on technology and the potential erosion of professional judgement. These risks do not necessarily originate solely in the algorithm. They can also arise from how a system is introduced, understood and used by people working under real clinical conditions.

Professional regulation therefore becomes an important part of implementation. A clinician using an AI recommendation remains part of a regulated profession with existing duties around competence, judgement, communication and patient safety. Regulators now need to determine how those established responsibilities translate when part of the information used in a decision is generated or interpreted by an AI system.

The organisations have not yet published the final principles. Their statement is a commitment to develop them jointly, meaning the practical requirements for individual professions still need to be worked out. The significance at this stage is the attempt to establish consistency before separate professional rules develop independently.

AI governance cannot stop at the algorithm

The UK initiative comes as the country develops a wider regulatory framework for healthcare AI. Earlier this month, the National Commission into the Regulation of AI in Healthcare recommended a more proportionate, lifecycle-based and system-wide approach. Its work extends beyond the regulation of AI-enabled medical devices to questions including accountability, transparency, clinical practice, organisational governance and assurance across the healthcare system.

That wider perspective is important because safe implementation depends on several layers operating together. Developers need requirements for designing and validating technologies, healthcare organisations need governance for introducing them into clinical workflows, and professionals need clear expectations for using them responsibly. Patients, meanwhile, need confidence that introducing AI does not make accountability disappear between the technology supplier, healthcare organisation and clinician.

The new collaboration among professional regulators addresses one of those layers directly. It recognises that AI competence may increasingly become part of professional competence itself. Healthcare workers do not necessarily need to understand how every algorithm is engineered, but they may need to understand enough about a system's purpose, limitations and reliability to decide when its output can inform care and when it should be challenged or ignored.

This also raises questions for education and workforce development. If AI becomes routinely embedded in healthcare, professional standards, continuing education and potentially training curricula will have to evolve alongside the technology. Safe adoption is therefore not simply an IT or procurement challenge; it increasingly becomes a workforce issue.

A model other countries will be watching

The UK's attempt to coordinate 38 regulators and registers could be relevant well beyond the country itself. Health systems internationally are confronting similar questions as AI moves from pilots into routine clinical workflows. European countries in particular are simultaneously implementing new AI regulation, expanding access to health data and trying to accelerate adoption, creating a growing need to translate system-level rules into practical expectations for healthcare professionals.

The difficult question will be how detailed those expectations should become. Principles need to be clear enough to protect patients and guide professionals without becoming so prescriptive that they cannot accommodate rapidly changing technologies or differences between clinical settings. A radiologist using an AI-enabled imaging system faces different risks from a nurse using an automated documentation tool or a psychologist encountering generative AI in administrative work.

The UK regulators are therefore addressing a part of healthcare AI governance that is likely to become increasingly important as adoption grows. Regulating algorithms remains necessary, but technology does not make clinical decisions in isolation. Ultimately, healthcare organisations and professionals must decide how AI is incorporated into care, when humans intervene and who remains accountable.

As AI moves deeper into everyday healthcare, those questions are becoming less theoretical. The next phase of AI governance will increasingly be about the people using the technology, not just the technology itself.