MHRA expands MedSafetyWeek focus to AI device safety and student reporting
The MHRA has tied its 2026 MedSafetyWeek campaign to a concrete educational push: a national patient safety essay competition on the detection and reporting of AI-related adverse events. The move reflects a regulatory problem that is arriving faster than many reporting systems were built to handle. As AI-enabled medical devices move deeper into routine care, the quality of post-market surveillance will depend on whether clinicians, patients, and trainees recognize small failures early enough to report them.
The campaign message is blunt: “Don’t ignore suspected side effects, report them.” It is aimed not only at medicines, vaccines, and traditional devices, but also at AI-enabled tools whose problems may look different from classic drug toxicities or mechanical malfunctions. Some adverse events may emerge as unexpected outputs, inconsistent recommendations, performance drift, or other behavior that does not match the clinician’s expectation of how a device should work. In practical terms, this means surveillance has to catch more than obvious breakdowns.
The competition itself is an attempt to move that thinking into the next generation of prescribers and device users. This year’s entry pool has been widened to include pharmacy students alongside UK medical students after strong interest from pharmacists who wanted to participate. Last year’s contest drew 68 entries from more than 30 medical schools, a response that shows the topic already resonates with trainees who will soon be handling medicines and devices at the bedside. Six prizes are on offer, with first prize receiving £500 and an invitation to present the essay at a national meeting of the Commission on Human Medicines and the Interim Devices Working Group.
The essay prompt is highly specific. Entrants are being asked to explain how the Yellow Card scheme can help identify risks linked to AI as a medical device, propose practical ways to improve detection and reporting of AI-related adverse events by patients and healthcare professionals, and discuss how stronger reporting could support safer use of AI-enabled devices. The framing matters because it pushes students beyond theory. It asks them to think about the reporting chain itself: how a patient notices something odd, how a clinician decides it is reportable, and how regulators interpret patterns that may only become visible after dozens or hundreds of individual reports.
Yellow Card reporting is being positioned as a frontline safety tool
The MHRA is using MedSafetyWeek to reinforce a simple point: no one needs to prove causality before submitting a Yellow Card report. Suspected side effects, device incidents, defective medicines, and concerns about fake or unlicensed products can all be reported when there is a reasonable concern. The threshold is intentionally low because safety surveillance depends on signals, not certainty. A single report may not prove much on its own, but it can be the first thread in a pattern that later changes prescribing, labeling, software updates, or device oversight.
That logic is especially important for AI-related harms. Unlike a predictable pharmacologic adverse effect, AI-related problems may be intermittent, context-dependent, or affected by changes in the model, the data it sees, or the way clinicians use its output. A device may appear acceptable in ordinary cases yet behave unexpectedly in unusual scenarios. Reports from the field are often the only way a regulator can identify those weak signals before they become normalized as “just how the system works.”
Alison Cave, the MHRA’s chief safety officer, stressed that every Yellow Card report adds real-world evidence about how medicines and devices are being used. She also singled out medical devices, including AI-enabled products, because reporting rates remain low. In regulatory terms, low reporting is not a minor administrative issue; it creates blind spots. If fewer frontline users submit reports, the agency has less chance of spotting trends early enough to protect patients.
A broader safety culture is being asked of patients, carers, and clinicians
The agency is not limiting the campaign to professionals. Patients, parents, carers, and healthcare workers are all being asked to report problems involving medicines, vaccines, herbal remedies, homeopathic products, e-cigarette products, medical device incidents, defective medicines, and suspected fake or unlicensed products. The message is deliberately inclusive because many safety events are first noticed outside the hospital or clinic. A family member may detect a new symptom after a medicine change. A patient may see a recurring device fault that never reaches the prescribing team. A pharmacist may spot a quality concern before it is linked to wider harm.
Reporting also does not require a fully detailed case file. The MHRA says reports remain valuable even when every detail is unavailable. That lowers the barrier for entry and encourages earlier submission, which is often when the information is most useful. In pharmacovigilance and device surveillance, timeliness can matter more than completeness. A partial report filed now can be more valuable than a polished report submitted after the patient has forgotten the sequence of events or the device has already been withdrawn.
Public awareness work around MedSafetyWeek is also being supported with campaign materials, including posters, digital screens, waiting room materials, and social media assets, alongside the hashtags #MHRAYellowCard, #MedSafetyWeek, #ReportSideEffects, and #patientsafety. The global reach is broad: more than 130 organisations from over 120 countries will share safety messages in more than 60 languages during the 2–8 November campaign window.
The regulatory signal is plain: AI safety will depend on ordinary reporting habits
The MHRA’s larger message is not about a single essay prize. It is about building a reporting culture before AI-enabled devices become more deeply embedded in routine clinical practice. If clinicians treat unusual AI behavior as a nuisance rather than a reportable event, safety surveillance will lag behind technology adoption. If patients and carers assume a symptom is too vague to mention, the system loses another possible signal.
For now, the competition deadline is 11 January 2027, and the topic is narrow by design. Entrants must link AI, device safety, and the Yellow Card scheme in a way that is clinically realistic and operationally useful. The unanswered issue is how quickly such educational efforts can translate into higher-quality reporting once AI products become more common across UK care settings. Until then, the MHRA is betting that better awareness, lower reporting thresholds, and a wider trainee audience will improve the safety net before the next device problem appears.