UK Proposes L-Plates for AI in Healthcare Regulation

New artificial intelligence tools used in healthcare settings should initially receive provisional, learner-style approval, comparable to an “L-plates” system. Developers would need to keep that status until they prove their tools perform safely and effectively in everyday clinical use. That’s the central recommendation from the UK’s largest-ever consultation on regulating health technology.

Continuous Monitoring, Greater Transparency

Published this week, the recommendations call for ongoing oversight of AI tools even after they’ve cleared initial approval. They also push for improved public access to safety-incident data. This comes as regulators work to keep pace with generative AI systems that keep evolving long after hospitals and clinics deploy them.

The UK’s Medicines and Healthcare products Regulatory Agency (MHRA) set up the initiative last year, known as the National Commission into the Regulation of AI in Healthcare, to build a regulatory approach suited to this new generation of AI-driven tools.

Two NHS doctors led the effort, and its findings drew on input from roughly 12,000 people, spanning patients, clinicians, health system leaders, and technology developers. The MHRA has indicated it will review the recommendations and respond formally in the near future.

Why Regulators Are Rethinking Their Approach

According to MHRA chief executive Lawrence Tallon, the agency wants the UK to be a leading destination for AI innovation in healthcare, a place where developers can build and test new tools, clinicians can adopt them with confidence, and patients ultimately benefit.

Tallon explained that AI-based medical tools present a unique regulatory challenge compared to traditional devices. Unlike older devices, they keep adapting after receiving approval as they’re exposed to more real-world data. That ongoing evolution can bring improvements. But it can also introduce what’s known as “drift,” where a system’s behavior gradually shifts away from its intended function. Either way, he said, regulators need a way to track what’s happening with these tools well after they’ve launched.

Jennifer Dixon, head of the Health Foundation, struck a more cautious note. Her organization contributed public-attitude research that helped shape the Commission’s recommendations. She pointed out that success will ultimately hinge on the NHS itself. The health service needs the resources, expertise, and infrastructure to roll out and monitor AI tools safely at the scale now required.

Looking to International Models

As part of its research, the Commission studied regulatory frameworks that the United States, Canada, and Japan are already developing. Among its key proposals is a publicly searchable database tracking safety information for AI-enabled medical devices, including records of adverse events.

Should Patients Be Told When AI Is Used in Their Care?

Research cited by the Commission found that most survey participants felt patients should be informed whenever AI played a role in their treatment. Currently, there’s no formal rule requiring healthcare providers to routinely disclose this information. The report acknowledges that flagging every single instance of AI use may not always be feasible. Still, it argues that clear, proportionate communication remains critical. This kind of transparency helps preserve trust between patients and the healthcare system.

The Commission also pushed for expanded enforcement powers for the MHRA, enabling the regulator to respond swiftly and firmly when AI systems fail to meet required standards.

Conclusion

Part of the urgency stems from a looming global workforce crisis. The World Health Organization projects a shortfall of 11 million healthcare workers by the end of the decade. Simply hiring more staff isn’t a realistic fix, Tallon argued. Technology and AI, he said, need to help the existing workforce deliver faster, more accurate care.