The UK’s Medicines and Healthcare products Regulatory Agency (MHRA) has launched three government-backed projects that embed artificial intelligence (AI) into the medicines pipeline, aiming to predict harmful side effects from drug combinations before treatments reach patients and to streamline regulatory decisions. The announcement, published on 22 October 2025, sets out more than £2 million in new funding to modernise how drugs are tested and approved without lowering safety standards.
Why this matters: polypharmacy is common—and risky
Millions in the UK take multiple medicines daily; in England, roughly 1 in 7 people are regularly prescribed five or more drugs. While most combinations are safe, some interact in ways that trigger side effects that can lead to repeat GP visits, prescription changes, or hospital admissions—burdens felt by patients, carers, and the NHS alike.
Project 1: spotting dangerous drug-drug interactions from real-world data
The flagship study will use AI to detect early “signals” of adverse interactions between medicines by analysing anonymised NHS data, starting with cardiovascular drugs. Led by MHRA scientists in partnership with PhaSER Biomedical and the University of Dundee, the work is funded with £859,650 from the government’s AI Capability Fund, coordinated by the Regulatory Innovation Office (RIO). The goal: inform prescribing decisions before risky combinations are widely used, preventing avoidable harm.
From data to lab bench: validating the signals in human-relevant models
Crucially, any AI-flagged risks won’t stop at pattern detection. The team plans to test signals in lab systems that mimic human drug metabolism and handling, providing biological confirmation that a predicted interaction is real and clinically meaningful. That loop—from NHS data to wet-lab validation—aims to raise confidence among clinicians and regulators that AI predictions translate into genuine safety improvements.
Potential impact: fewer admissions, faster answers
Adverse reactions to medicines are estimated to account for around one in six hospital admissions in England and cost the NHS more than £2 billion annually. Regulators say using AI and real-world data could help prevent a portion of these events by identifying hazardous combinations sooner, improving clinical guidance and prescribing safety.
De-risking development: catching failures earlier saves time and money
Late-stage drug failures remain stubbornly high across the industry. By analysing diverse, real-world health data, AI tools could flag safety and efficacy issues earlier, reducing costly late-stage surprises and giving regulators stronger evidence for faster, well-informed decisions. Industry analyses have long highlighted the challenge of late-stage attrition, underscoring the value of better predictive tools.
Voices from the regulator and partners
MHRA Chief Executive Lawrence Tallon framed the initiative as a core part of modernising regulation to match how people live now—often managing multiple conditions with multiple medicines—while keeping safety central. Project supervisor Julian Beach said the work is a step toward “smarter” trial design that builds AI and advanced modelling into the process. PhaSER’s CEO Chris Wardhaugh said the collaboration brings a “human-relevant lens” on metabolism and safety into regulatory practice.
Project 2: AI assistants for regulatory insight, safety and efficiency (ARISE)
A second MHRA programme—AI for Regulatory Insight, Safety, and Efficiency (ARISE)—will pilot AI-assisted tools to support expert reviewers in scientific advice, clinical trial assessment, and licensing. Backed by £1,000,000 via the Regulators’ Pioneer Fund, the system is designed to improve consistency and throughput while leaving final decisions with humans, aligning with the Life Sciences Sector Plan’s push for faster, risk-proportionate approvals.
Project 3: testing synthetic data to strengthen evidence where patients are few
The third project (£259,250, also via the Regulators’ Pioneer Fund) explores synthetic patient data—artificially generated datasets that statistically resemble real data—to supplement clinical trials in areas like cancer, inflammatory bowel disease, and rare paediatric seizure conditions. The work will run in a regulatory “sandbox,” probing when and how synthetic data can responsibly bolster evidence where recruiting large cohorts is difficult.
Part of a wider push to cut red tape and unlock innovation
The funding sits within a broader government drive to help regulators pilot new, pro-innovation approaches so that promising technologies can reach people faster without compromising safety. On the same day, the Department for Science, Innovation and Technology highlighted 16 projects backed through the RIO to reduce regulatory friction across sectors—from drones delivering medicines in remote regions to AI tools that help inspectors assess risks more quickly.
How the pieces fit together: data, methods, and policy
Beyond specific tools, the MHRA projects will produce practical guidance for developers on combining AI, real-world data, and traditional trials to generate decision-grade evidence. That includes input to the MHRA’s National Commission into the Regulation of AI in Healthcare and alignment with ambitions to make the NHS one of the world’s most AI-enabled health systems. The intent is to build methods, rules, and expectations that are transparent and usable across the medicines ecosystem.
Guardrails: human oversight and inclusivity by design
Officials stress that all final regulatory decisions will remain with experts, who will use AI as decision support—not decision replacement. Using real-world NHS data is also meant to reflect patient diversity more accurately than some clinical trials, potentially surfacing safety issues that might otherwise be missed. The projects’ design—combining population-scale data with human-relevant lab models—aims to balance innovation with caution.
What success could look like for patients and clinicians
If the approach works, prescribers could get clearer, earlier warnings about dangerous combinations, personalised to patient context, while developers gain earlier readouts on risk. That could reduce trial amendments, speed safe approvals, and prevent some hospital admissions linked to adverse interactions—all while supporting the UK’s positioning as a global life-sciences hub.
Key numbers at a glance
The package totals over £2 million in new support for the MHRA: £859,650 for AI-driven detection and validation of drug-drug interactions; £1,000,000 for the ARISE programme to pilot AI assistants in regulatory workflows; and £259,250 to test synthetic data as a complement to trial evidence. These investments align with a separate £8.9 million tranche supporting 16 cross-sector regulatory pilots coordinated by the RIO.
The MHRA’s move signals a practical turn in the AI-for-health conversation: less hype, more tooling that answers immediate safety and efficiency needs. By pairing NHS-scale data with lab validation and keeping humans in the loop, the regulator is betting it can both reduce harm from drug interactions and accelerate the delivery of effective treatments—two goals that, with the right safeguards, can reinforce each other.