MHRA's Record Annual Report Masks a Governance Gap Closing In on UK Healthtech

The MHRA does not often make headlines for sounding optimistic. So when its 2025-26 Annual Report, laid before Parliament on 15 July, describes a year of record delivery on every statutory target, it repays careful reading.

What the Report Actually Says

The headline figures are striking. The agency approved 921 medicinal products during the year, including 39 new medicines, and assessed 100% of clinical trial and national licence applications within statutory timelines. It also removed a record 28.5 million doses of illegally traded medicines from circulation, with an estimated street value of £64.4 million. For the healthtech sector, the most significant passages concern AI.

The report confirms that the MHRA played a leading role in shaping future AI regulation, including the launch of the National Commission into the Regulation of AI in Healthcare, an advisory body drawing on global AI leaders, clinicians and regulators, to inform a dedicated regulatory framework for AI as a Medical Device (AIaMD). The agency also expanded its AI Airlock programme, the UK's first regulatory sandbox for AIaMD, which has now completed Phase 2 and is being developed into a third cohort. Phase 3 focuses on some of the hardest technical terrain: large language models in clinical decision support, synthetic data for radiology validation, and real-time post-market surveillance.

CEO Lawrence Tallon described the year as one of "delivering faster, more consistently and with even greater impact for patients, the NHS and innovators." An independent customer survey found that 83% of respondents felt the MHRA was doing a good job regulating medicines, up eight points year-on-year.

The Sandbox Is Only Part of the Story

The AI Airlock is a genuine regulatory innovation. Sandboxes, by design, serve only a small cohort of companies at the frontier. The question for the broader market is whether the hundreds of smaller operators building AI tools for clinical and care settings are keeping pace with what regulators will eventually expect of them.

The answer, at least according to new data from Birdie, is: not quite. The homecare technology company's July 2026 report, Moving Faster than the Rules: AI, Care Quality and the Homecare Sector in 2026, surveyed 122 homecare providers across the UK and found that 70% are already using AI in some form, rising to a projected 85% within a year. Around half are using it not just for administration but to directly shape care plans and risk assessments. Among providers re-inspected since adopting AI, 59% saw their CQC rating improve, and none reported a decline.

The governance picture is uncomfortable. Only 66% of AI-using providers have any formal policy governing how they use it. Just 43% have a written policy in place at all.

Brand Exposure the Sector Has Not Noticed

Governance policies are not the only thing missing. AI Business Dispatch analysis of Companies House and UK Intellectual Property Office data (as of July 2026) reveals a separate vulnerability hiding in plain sight.

Just 60 new companies registered under SIC code 86.10 (Hospital Activities) were incorporated in Q3 2026, a drop of 80.8% versus the prior period, suggesting the pipeline of new hospital-adjacent operators is compressing sharply. Meanwhile, UK trademark filings in Nice Class 44 (medical and healthcare services) fell to 531 in the same quarter, down 76.9% on the prior period. Most telling: 98.3% of active SIC 86.10 companies hold no Class 44 trademark whatsoever.

That figure should concentrate minds. In a market where NHS procurement teams, private equity backers, and now the MHRA itself are demanding evidence of sound governance and legitimate market presence, operating a healthcare AI brand without registered trademark protection is an avoidable liability. Class 44 covers medical services directly, the natural home for any brand offering clinical AI tools or care-adjacent software. The gap between incorporation activity and trademark activity is not a technicality; it is a measure of how many operators are building without a legal foundation for their brand.

The NHS Accelerates Anyway

Above all this sits a far larger engine. NHS England's 4 July announcement confirmed a major AI rollout backed by £10 billion in technology investment over the next three years. A new AI triage tool in the NHS App, which showed a 29% reduction in phone queuing in a Sussex GP trial, is set to reach more than 200,000 patients within 12 months and all NHS App users by April 2028. Separately, 505,000 clinicians and support staff are being given access to Microsoft 365 Copilot, with NHS England estimating average savings of around 43 minutes of administrative time per day per worker.

The NHS also plans an England-wide rollout of AI ambient scribing tools, recording consultations, generating real-time transcriptions and clinical summaries. The 10 Year Health Plan explicitly aims to make the NHS "the most AI-enabled care system in the world."

The regulator is building governance infrastructure. The NHS is deploying at scale. Homecare providers are already using tools their own written policies have not caught up with.

The Gap Regulators Will Close

The MHRA's forthcoming dedicated AI as a Medical Device framework, expected later in 2026 following the draft Medical Devices (Amendment) Regulations published in May, will formalise requirements that the market has so far been handling informally. Predetermined Change Control Plans for adaptive AI, stricter post-market surveillance, and an International Reliance Pathway for cross-border approvals are all coming into force.

For operators who have moved fast and not yet documented their governance, the window for comfortable catch-up is narrowing. For those who have not yet registered a trademark in Class 44, it is arguably already overdue. The MHRA's record year marks how far the regulatory frontier has advanced, and how far most of the market still has to travel to meet it.