UK AI regulation in healthcare: MHRA's 44 recommendations signal new compliance rules for builders

The UK's Medicines and Healthcare Products Regulatory Agency (MHRA) has published 44 recommendations to overhaul how AI products are regulated in the NHS and other healthcare settings. The proposals include continuous monitoring of AI after approval, mandatory patient transparency, a supervised 'L plate' system for trialing new models, and the power to penalise developers whose products fail to meet standards. For builders shipping AI into healthcare, this signals a shift from static pre-market approval to ongoing lifecycle oversight.

MHRA's 44 recommendations: a new regulatory framework for AI in healthcare

The MHRA, which regulates all medical devices and licenses treatment drugs in the UK, says the current framework was designed for products like hip replacements and stethoscopes, not for AI systems that "continue to change after the point of authorization" as new data is fed in. An independent commission gathered input from more than 12,000 patients and clinicians to shape the recommendations.

Key proposals include:

• Continuously monitoring AI products and removing them from regulatory approval if they malfunction or become less effective over time
• Giving patients the right to know whether AI is involved in their care and easy access to information about the products
• The power to penalise developers if an AI product fails to meet required standards
• An AI 'L plate' system to allow new models to be trialled by healthcare professionals under close supervision

MHRA chief Lawrence Tallon told the BBC that patients will "increasingly see AI as part of the way that normal NHS healthcare is delivered" and that the system must maintain trust and confidence.

Why AI builders should prepare for post-authorization oversight

The most consequential shift for developers is the emphasis on post-authorization learning. Unlike a hip replacement, an AI model can drift as it encounters new data. The MHRA explicitly acknowledges this: "As new data gets fed in, they learn, they adapt, they drift." Builders will need to design for continuous data collection, model versioning, and safety reporting after deployment. The ability to remove a product from approval if it underperforms means ongoing validation is not optional.

Transparency requirements also affect product design. Patients must know when AI is involved in their care and have easy access to information about the product. This will influence consent flows, user interfaces, and how AI decisions are explained.

Practical implications: AI scribes, L plates, and patient trust

AI note-takers known as scribes, powered by LLMs, are reportedly used by 40% of UK GPs to record consultations and generate reports. A University of Edinburgh study found patients may be less likely to share personal information such as substance abuse history if they know the conversation is being processed by AI. The proposed regulations would formalise transparency around such tools.

The 'L plate' system is a practical mechanism for supervised trialing. It could lower the barrier for testing novel AI tools in clinical settings while maintaining oversight. However, it also implies a staged rollout process that developers must factor into their go-to-market plans.

Caveats: bias, global coordination, and unfinished policy

The recommendations are not yet law. They represent regulatory intent, and actual rules may evolve. Tallon acknowledged that no country has "absolutely cracked" AI regulation, and global coordination remains a challenge.

Concerns persist about biases in training data and the risk of incorrect medical advice from AI chatbots. The MHRA's framework will need to address these through validation requirements and human oversight. Builders should also watch for how the UK's approach aligns with EU AI Act and FDA frameworks, as multi-jurisdiction compliance may become complex.