MHRA and NICE Use DMHT Working Group to Set the Regulatory Ground Rules for Digital Mental Health
The MHRA’s Digital Mental Health Technology Working Group became a practical engine for shaping how digital mental health technologies are regulated, evaluated, and eventually tested against evidence standards in the UK. Established in phase 1 of a Wellcome-funded project, the group brought together regulatory experts, health technology assessment specialists, NHS policy voices, clinicians, researchers, and people with lived experience to address a problem that has become impossible to ignore: software products used for mental health care often move faster than the systems meant to assess them.
The timetable itself tells the story. The first meeting took place in September 2023, with monthly sessions at the outset and then meetings every other month until April 2026. Over that span, the group moved from broad conceptual discussion to detailed review of draft guidance, evidence needs, adverse event reporting, and the emerging role of artificial intelligence in mental health software. In regulatory terms, this is not a passive advisory panel. It is a mechanism for translating multidisciplinary judgment into guidance that manufacturers, evaluators, and health systems can actually use.
From Conceptual Debate to Device Characterisation and Classification
Early discussions focused on the core question that sits beneath nearly every dispute about digital mental health products: what counts as a device, how should it be characterised, and when does it qualify as software as a medical device (SaMD)? The April 2024 in-person meeting marked a turning point when MHRA presented a framework for device characterisation, regulatory qualification, and classification. A follow-up meeting in May 2024 narrowed the discussion further and helped reach a more focused consensus.
This matters because digital mental health tools do not sit neatly in a single category. Some products function like educational tools, others resemble symptom trackers, and some deliver interventions that can affect diagnosis, monitoring, or treatment decisions. When software is used to influence care, regulators need a clear line between consumer-facing wellness content and tools that may alter clinical management. The working group’s feedback fed directly into MHRA guidance on qualification and classification, giving the project a more usable regulatory spine.
Quality of Life, Evidence Standards, and the Problem of Measuring Benefit
By June 2024, NICE had shifted the discussion toward a question that routinely complicates assessment of mental health interventions: how should quality of life be measured for DMHTs? Mental health technologies often promise benefits that are broader than symptom reduction. They may improve functioning, support engagement, or offer access in settings where face-to-face care is limited. Yet these effects are difficult to capture if evaluators rely on a narrow outcome set.
That difficulty has direct implications for evidence generation. If the wrong outcome measure is chosen, a useful product can appear ineffective, while another with modest clinical value can seem persuasive because it performs well on a surrogate endpoint. The working group’s discussion fed into subsequent research on the impact of health-related quality of life measures in mental health and helped shape later work on standardising outcome collection in clinical studies. In effect, the group was helping define what counts as meaningful evidence before the evidence pipeline was fully built.
A digital intervention is only as credible as the outcome measures used to judge it.
App Stores, Adverse Events, and the Safety Questions Manufacturers Cannot Avoid
The working group also tackled the less glamorous but essential mechanics of the DMHT ecosystem. December 2023 sessions examined app store responsibilities, while later meetings explored the wider environment in which these products are discovered, downloaded, and used. That discussion is important because app stores are not neutral shelves. They shape visibility, influence user expectations, and can blur the line between regulated medical software and general consumer health content.
Safety became more explicit in April 2025, when MHRA discussed adverse event reporting and potential phase 2 topics. The group’s input led to more DMHT examples in MHRA guidance for manufacturers on reporting adverse incidents involving software as a medical device. This is clinically relevant because digital products can generate harms that differ from those seen with pills or procedures. A software update can alter functionality overnight, a poorly designed interface can mislead users, and a generative AI feature can amplify unsafe advice or unstable behavior. If regulators do not ask the right questions, these failures can remain invisible until harm has already occurred in routine use.
AI Airlock Priorities and the Shift Into Phase 2
The working group’s influence did not stop at guidance documents. It helped shape the Digital Mental Health Technology Artificial Intelligence AI Airlock, one of the defining phase 2 priorities. In April 2026, MHRA and NICE presented phase 2 work packages and Delphi results that informed AI Airlock priorities and the call for applications. That sequence reflects a broader regulatory reality: generative AI has introduced fresh uncertainty into mental health care, where language, context, and risk assessment are inseparable from safety.
For mental health software, AI is not simply a technical upgrade. It changes the clinical risk profile. A system that answers users, triages symptoms, or generates recommendations may appear helpful while introducing unpredictable failure modes. The working group’s December 2025 discussion on the risks of generative AI in mental health fed directly into MHRA thinking. Regulators are therefore moving toward a model in which innovation is not blocked, but routed through more explicit scrutiny of evidence, safety, and intended use.
A Multidisciplinary Panel as Regulatory Infrastructure
The strength of the working group lay in its composition. MHRA and NICE experts sat alongside NHS England policy voices, psychiatrists, clinical psychologists, mental health nurses, GPs, academic researchers, and lived experience advisors. That mix matters in mental health because technical safety, patient acceptability, and service delivery often collide in the same product. A tool can be methodologically sound and still fail if it ignores workflow, literacy, accessibility, or the realities of people using it during periods of distress.
This breadth also explains why the group’s discussions shaped so many downstream outputs: the user and public perspectives research plan, the regulation and evaluation publication, the conceptual framework paper, draft guidance, e-learning for patients and professionals, and later evidence-generation work with IMDRF and BSI. The meetings served as a steady filter, turning broad policy intent into concrete regulatory and educational material.
What the Record Shows About UK Digital Mental Health Policy
The main achievement here is cumulative. Over nearly three years of meetings, the working group helped MHRA and NICE refine how DMHTs are defined, assessed, measured, and monitored. It also pushed the system to confront questions that have no easy shortcuts: what counts as acceptable evidence, how to handle software-related harm, which outcome measures matter most, and how to govern AI before its clinical use outpaces oversight.
The document closes with an acknowledgment of the group members who contributed to that effort, a reminder that the regulatory framework for digital mental health has been built through repeated expert negotiation rather than a single decisive announcement. The unresolved challenge is equally clear: as generative AI features spread through mental health software, the next phase of regulation will have to test whether current qualification rules, evidence standards, and post-market reporting systems are strict enough for products that can change after deployment.