The UK wants to build the "most AI-enabled healthcare system in the world." That ambition is noble. The problem is that the regulations meant to keep that AI safe and effective are, by the industry's own assessment, fundamentally broken.
A new report from the MHRA's National Commission on the Regulation of AI in Healthcare — based on responses from 761 stakeholders including NHS bodies, industry, academia, and patient groups — delivers a damning verdict on the current framework. Seven in ten respondents disagreed or strongly disagreed that existing regulations are sufficient on safety and performance standards. Six in ten said the same about data privacy and governance. More than half want either substantial revision or a complete overhaul.
This isn't nitpicking. These are the people actually building, deploying, and working with AI systems in clinical settings saying the rules don't work.
What's Wrong With The Current Setup
Right now, AI in UK healthcare falls under the medical device regulation (MDR). That means AI-powered diagnostic tools, clinical decision support systems, and predictive algorithms get lumped in with hardware medical devices. The problem is that software doesn't behave like a pacemaker. An AI model can drift, learn, and change behaviour in ways a physical device simply cannot.
When your AI system that was approved in 2024 starts performing differently in 2026 because of updates or new training data, the MDR has no clear mechanism for handling that. Post-market surveillance — the process of monitoring a device after it's been deployed — was described as inadequate by 65% of respondents. If an AI system starts making worse decisions over time, who catches it, and how?
The other gaps are equally stark. Transparency requirements are insufficient according to 71% of respondents. Clinical evidence standards don't account for how AI systems learn and adapt. Data governance frameworks weren't designed with AI in mind at all.
What The Industry Wants
The respondents — and remember, these include the companies building AI health products, not just critics — are calling for clearer responsibilities across the AI lifecycle. They want comprehensive governance structures that oversee development, deployment, and ongoing monitoring. They want healthcare professionals trained properly to use these systems. They want confirmation that AI tools actually address unmet clinical needs rather than just being novel.
What's striking is the constructive tone. This isn't an industry trying to escape regulation. It's an industry drowning in the wrong kind of regulation, asking for the right kind.
The National Commission's final recommendations are expected in the coming months. If the government actually listens — and if the promised alignment with the NHS 10-year plan and Life Sciences Sector Plan materialises — the UK could actually build a regulatory framework that doesn't force companies to choose between innovation and safety.
That's the hope, anyway. We'll see if the execution matches the ambition.
Top comments (0)