UK medicines regulator takes its AI device rulebook to the US, seeking FDA ties

The MHRA used a Boston medtech conference to outline lifecycle regulation, sandboxes and closer work with the FDA for AI-enabled medical devices. The UK has also accepted 44 expert recommendations.

UK medicines regulator takes its AI device rulebook to the US, seeking FDA ties

Britain’s medicines and devices regulator is trying to write the rules for AI in medicine before the technology outruns them. It also wants the US Food and Drug Administration working alongside it.

Jessica Paulsen, the Medicines and Healthcare products Regulatory Agency’s deputy director for AI and software, set out the plan on 20 October. She was speaking at the AdvaMed medtech conference in Boston. The plan has since been picked up in legal and industry round-ups, including TLT’s November AI brief. This article rests on the regulator’s announcement as reported by trade outlets, not on the full text of her talk.

Advertisement

From old rules to lifecycle oversight

Medical device regulation was built mainly for products that stay the same once approved. AI software often does not. It can be updated, retrained and applied in new settings. According to ICT&health, the MHRA therefore wants a more proportionate, lifecycle-based approach, with safety checked continuously and not at a single approval point.

The ideas sit inside the MHRA 2030 Strategy. That strategy aims to make the agency a source of a global standard for safe and agile regulation of medtech and AI by 2030, with faster access to new technologies and stronger patient safety, as reported by HTWorld.

Paulsen’s central message, as quoted by several outlets, was that AI is changing medical technology and regulation must keep pace. Resultsense reports she also said the UK is “moving from policy into practice”.

Advertisement

What is actually planned

The reported measures include:

  • A new MHRA Centre for AI and Digital Health, intended to pool regulatory expertise and give developers clearer guidance, according to ICT&health.
  • Continued use of AI Airlock, a regulatory sandbox where developers and regulators test problems posed by AI-enabled devices in a controlled setting.
  • Wider device reform, including better post-market surveillance and proposed legislative changes to allow a more responsive framework, as reported by Pharmacy Business.

SelectScience says AI Airlock is backed by partnerships with NHS London and Manchester University NHS Foundation Trust. The aim is to see how devices perform in real clinical settings and to produce evidence for proportionate regulation. One law firm update, from Arqon, adds that the MHRA opened applications on 6 October for a third phase of the sandbox. That phase focuses on post-market surveillance and lifecycle oversight.

The expert commission

The government has also accepted all 44 recommendations of the independent National Commission into the Regulation of AI in Healthcare. HTWorld describes the commission as NHS doctor-led. Its recommendations support the same lifecycle-based model the MHRA is now describing. The reports reviewed do not list the 44 recommendations, so their detail is not covered here.

Advertisement

Why the FDA matters

Most AI medical devices are sold internationally, and a manufacturer facing two regulators has to satisfy both. Paulsen said, in Resultsense’s account, that because these are global technologies, collaboration with the FDA “will be critical”.

The MHRA is recruiting a Washington-based liaison officer to work with the US agency, according to Resultsense. Pharmacy Business describes the programme as meant to deepen day-to-day cooperation, scientific exchange and coordination on emerging problems, including medical devices and AI.

The idea is not new. HTWorld notes that the MHRA set out ambitions at AdvaMed in 2025 to align more closely with the FDA, ease transatlantic market access and make wider use of international reliance routes. Reliance means leaning on another regulator’s approval, and it would cover devices the FDA has already cleared.

Advertisement

Citeline’s Medtech Insight, reporting from the same conference, said the FDA is preparing draft guidance on generative AI in devices. It said the MHRA is weighing reforms such as predetermined change control plans and stronger post-market monitoring. Those are plans, not final rules.

What remains open

The announcements describe direction, not binding law. The proposed legislative changes have not been passed. The new AI centre is planned, not yet operating, in the sources reviewed. No source gave a timetable for the FDA liaison role or for when the lifecycle rules would take effect.

Developers will want to know how a continuously updated product is supposed to stay compliant. Hospitals will want to know who is accountable when a model changes after purchase. The MHRA’s answer, for now, is to test those questions in the sandbox before writing them into rules.

Advertisement

Sources