MHRA targets medicine safety with new AI sandbox
Decision brief
Answer first · skim in under a minute
MHRA targets medicine safety with a new AI sandbox designed to let innovators test AI systems under regulatory supervision. The programme is intended to improve safety, speed development, and reduce reliance on animal testing.
Key questions this brief answers
- What did the MHRA announce on 9 June 2026?
- How many AI approaches will the first sandbox phase test?
- How does the AI sandbox differ from the AI Airlock?
The MHRA targets medicine safety with a new AI sandbox announced on 9 June 2026. The UK regulator will let innovators test AI tools that predict how medicines behave in the body under regulatory supervision, aiming to cut adverse reactions, speed development, and reduce animal testing.
Contents9 sections
Key Takeaways
- On 9 June 2026, the MHRA launched an AI sandbox for medicines safety and development, announced during London Tech Week.
- Adverse drug reactions send about 250,000 people to hospital in the UK each year and cost the NHS over £2 billion annually.
- Around 90% of drugs fail during development; the sandbox will test up to five AI-driven approaches in its first phase from summer 2026.
- Separately, the MHRA’s AI Airlock for AI as a medical device secured £1.2 million per year for 2026–2029 (£3.6 million over three years).
What did the MHRA announce for medicine safety?
Science Minister Lord Vallance announced the programme during London Tech Week on 9 June 2026. The Medicines and Healthcare products Regulatory Agency will create a controlled testing environment where companies and researchers work with regulators on AI tools for medicines safety and side-effect prediction.
According to the official MHRA press release on GOV.UK, the sandbox is funded through the UK Government’s Regulatory Innovation Office. It is designed to build an evidence base on whether AI tools can support decisions about the safety of new medicines.
Why does the MHRA want AI in medicines development?
The agency cites a clear safety and attrition problem. Adverse drug reactions send around 250,000 people to hospital in the UK every year, costing the NHS over £2 billion annually. Around 90% of drugs fail during development, in part because existing methods cannot always predict how medicines will behave in people.
The sandbox will explore AI tools that may predict absorption, processing, and harm potential earlier. It will also examine how better use of clinical data can improve understanding across groups often underrepresented in studies, including children, older people, and people from diverse backgrounds. Coverage of related EU regulatory news sits alongside Nova Pharma News tracking of UK and EU policy shifts.
How will the first sandbox phase work?
Up to five AI-driven approaches will be tested in the first phase. The MHRA said it will begin working with industry and academic partners from summer 2026 to shape how the sandbox operates.
The programme forms part of a wider modernisation push that includes advanced modelling and synthetic data. Officials link it to the Government’s AI for Science Mission One, alternatives to animal testing plans announced the prior November, and the 10 Year Health Plan goal of a highly AI-enabled healthcare system. Teams following EU regulatory category coverage can track how sandbox evidence may feed later guidance.
How does this relate to the MHRA AI Airlock?
The medicines AI sandbox is distinct from the AI Airlock, the MHRA’s regulatory sandbox for artificial intelligence as a medical device (AIaMD). On 8 April 2026, the MHRA said the Department of Health and Social Care allocated £1.2 million per year for 2026–2029 to expand Airlock after a successful second phase.
That uplift totals £3.6 million over three years and is meant to support longer-term testing beyond yearly budget cycles. Details are in the April 2026 Airlock funding announcement. Background on the programme is collected at the AI Airlock GOV.UK collection.
What should BD and regulatory teams watch next?
Near-term milestones are partner selection from summer 2026 and the first cohort of up to five AI approaches. Sponsors should track how MHRA defines acceptable evidence for AI pharmacokinetic and safety models before treating outputs as decision-grade.
Industry groups such as the BioIndustry Association and ABPI welcomed a supervised “safe harbour” for model testing. Separately, Airlock phase-two reporting is expected in summer 2026 and will inform phase three for AIaMD. Compare this UK path with broader AI in pharma EU coverage when mapping multi-region regulatory strategy.
What are the patient and NHS stakes?
Health Innovation Minister Preet Gill framed the sandbox as a way to cut adverse reactions, reduce animal testing, and speed safer treatments. MHRA Chief Executive Lawrence Tallon stressed stronger safety evidence that better reflects diverse patient populations.
If AI tools prove reliable under MHRA supervision, earlier risk detection could reduce late-stage failures that strand promising candidates. Until that evidence base exists, the sandbox remains a supervised learning environment, not a shortcut to market authorisation.
Frequently Asked Questions
What did the MHRA announce on 9 June 2026?
The MHRA announced a new regulatory AI sandbox to test tools that may improve medicines safety assessment, speed development, and reduce reliance on animal testing, with work starting from summer 2026.
How many AI approaches will the first sandbox phase test?
Up to five AI-driven approaches will be tested in the first phase, according to the MHRA press release.
How does the AI sandbox differ from the AI Airlock?
The June 2026 AI sandbox focuses on medicines development and safety prediction, while the AI Airlock is the MHRA’s regulatory sandbox for artificial intelligence as a medical device, funded at £1.2 million per year for 2026–2029.
