Europe Launches First Continent-Wide Expert Group to Steer AI in Mental Health

Munich has become the starting point for what many researchers regard as a pivotal moment in European psychiatry. The first Europe-wide expert working group dedicated to artificial intelligence in mental health has been launched at the 39th ECNP Congress, bringing together specialists who intend to map, evaluate and ultimately guide how AI technologies are developed, validated and deployed across the continent’s mental health systems. The initiative, established by the European College of Neuropsychopharmacology, arrives at a time when machine learning tools, large language models and automated chatbots are moving from research laboratories into clinical settings faster than regulators, clinicians and patients can fully assess their consequences. The group’s founders argue that without coordinated oversight, Europe risks a fragmented landscape in which promising innovations stall while unvalidated tools reach vulnerable people unchecked.

The launch was marked by an open discussion session at the congress, titled Artificial intelligence in mental health – Progress and perils, where the group set out its view of the field’s potential, its hazards and the directions it believes research and policy should now take. According to the organisers, the session was designed not as a closed expert briefing but as an invitation to the wider psychiatric and neuroscience community to scrutinise, challenge and refine the group’s plans. That consultative approach reflects a core conviction among the founders: that AI governance in mental health cannot be built by a small circle of technologists alone, but requires input from clinicians, patients, ethicists, regulators and industry alike.

Chairing the new group is Dr Paris Alexandros Lalousis of King’s College London, a researcher known for work on machine learning approaches to psychiatric diagnosis and prognosis. In remarks at the launch, he emphasised the pace at which AI is advancing and the absence of structures to guide it. AI, he noted, offers great opportunities in mental health, but its development also needs to be guided, communicated and regulated, something he described as currently missing. His stated ambition for the group is to produce a framework for understanding the risks and benefits of applying AI in mental health, complete with recommendations on how the field should move forward. He stressed that this is the first group to take on such a challenge at a European scale, and that a central early task will be mapping the field itself, so that Europe-wide resources exist to show researchers who is doing what and who is regulating AI across different jurisdictions.

The idea of mapping the field deserves particular attention, because it addresses one of the most persistent problems in psychiatric AI research: fragmentation. Across Europe, hundreds of teams are applying machine learning to questions ranging from predicting treatment response in depression to detecting early signs of psychosis from speech patterns, from drug discovery to the construction of therapeutic chatbots. Yet these efforts frequently proceed in isolation, using different datasets, different evaluation standards and different reporting conventions. The result, critics have long argued, is a literature full of models that look impressive in papers but fail when tested on new populations, and a regulatory environment in which no one has a complete picture of what is actually being deployed. By creating a shared map of who is working on what, and where regulatory responsibility lies, the group hopes to replace this patchwork with something closer to a coordinated European effort.

Co-chair Tim Hahn of the University of Muenster in Germany underlined both the breadth and the limits of the group’s expertise. The working group, he explained, comprises 19 experts drawn from Europe and the United States, whose research spans an unusually wide range: basic machine learning applied to drug discovery, patient concerns, fundamental neuroscience, and the building and evaluation of chatbots. The group has already developed an initial three-year work plan and is now meeting with the wider psychiatric and neuroscience community to gather feedback and refine it. Hahn was candid about the difficulty ahead, saying that even with such varied expertise, the biggest challenge will be to truly understand what people are doing right now and to determine where the group, as a network, should insert itself to provide guidance. His message to the community was deliberately inclusive: whether one views AI through a positive or a negative lens, the group wants people to engage with it.

The group has articulated a set of network outcomes that sketch the shape of its long-term ambitions. These include European evaluation standards for psychiatric AI, regulatory-aligned deployment guidance, shared methodological infrastructure, multisite validation studies, and training and capacity building. Each of these addresses a well-documented weakness in the current landscape. Evaluation standards are needed because psychiatric AI models are frequently reported without rigorous external validation, making it nearly impossible for clinicians to judge whether a tool will work in their own patient population. Deployment guidance aligned with regulation is intended to bridge the gap between the EU’s emerging AI governance framework and the practical realities of clinical practice. Shared infrastructure and multisite validation studies speak to the need for large, diverse datasets that can demonstrate whether models generalise across countries, health systems and demographic groups, a hurdle that single-centre studies routinely fail to clear.

In the medium term, the group has committed to a series of three-year deliverables that translate those outcomes into concrete outputs: a responsible AI position paper, an AI in Mental Health Roadmap, a living guideline for evaluation and reporting, training programmes across ECNP, and joint European funding initiatives. The notion of a living guideline is especially significant. Unlike a static set of recommendations, a living guideline would be continuously updated as the technology and the evidence base evolve, an approach increasingly favoured in fast-moving areas of medicine where fixed guidance risks obsolescence within months. Training programmes, meanwhile, acknowledge that the workforce itself must change: psychiatrists, psychologists and neuroscientists will increasingly need enough technical literacy to interrogate the AI tools they are asked to trust, and data scientists entering mental health need grounding in the clinical and ethical complexities of the patients they aim to help.

The launch has been framed within a broader strategic decision by the ECNP, Europe’s largest clinical neuroscience organisation, whose interests stretch from psychiatry through drug development, psychology and neuroscience. ECNP President Professor Andreas Reif of Frankfurt commented that AI has manifold implications for mental health, from research to risk factors to therapy, and that it will certainly change the way clinicians interact with their patients. For that reason, he said, ECNP has decided to make AI in mental health one of its priority fields, with the new AI Network serving as one of the key structures within that commitment. He expressed personal satisfaction that such an excellent international team could be assembled for the purpose. The president’s framing captures what makes this moment distinctive: AI is not being treated merely as a set of tools to be evaluated, but as a force that will reshape the therapeutic relationship itself, from how symptoms are measured to how support is delivered between appointments.

Independent voices have also welcomed the initiative while underlining why it is needed. Professor David Nutt of Imperial College London, who is not involved in the group, offered an external comment applauding the ECNP for orchestrating what he called a timely and much needed initiative, expressing his trust that it will bring balance and quality control to a rapidly expanding field that so far has little of either. His assessment echoes a widespread concern among researchers and clinicians: the commercial and academic incentives driving AI development in mental health have outpaced the mechanisms traditionally used to ensure safety and efficacy, leaving a gap that voluntary standards and professional networks are attempting to fill. Whether a professional society network can exert sufficient influence on developers and regulators remains an open question, but the group’s explicit intention to engage with regulation suggests it does not intend to remain purely advisory.

The group’s ambitions extend beyond the congress hall. At the Munich meeting, members also discussed acting as a resource for media enquiries relating to AI in mental health, a role that recognises how public understanding of these technologies is shaped by coverage that often swings between utopian promise and dystopian alarm. Immediately after the launch, Dr Lalousis was due to represent the ECNP AI group at a conference at the European Parliament, advocating for cross-border validation, regulation, implementation, workforce development and equitable access in precision psychiatry. That parliamentary engagement signals the group’s awareness that the hardest questions, such as who pays for validation studies, how tools are certified across member states, and how access is ensured for under-resourced health systems, are political as much as scientific. If the group succeeds in connecting Europe’s scattered expertise into a coherent, expert-driven framework, it could determine whether the continent’s mental health systems harness AI on their own terms or simply absorb whatever the global market delivers. The next three years of deliverables will show whether that ambition can be realised.

Subject of Research: Establishment of the first Europe-wide expert working group on artificial intelligence in mental health

Article Title: Launch of first Europe-wide expert group on AI in mental health

Article References: Launch of first Europe-wide expert group on AI in mental health. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: artificial intelligence, mental health, psychiatry, ECNP, machine learning, regulation, chatbots, precision psychiatry, clinical neuroscience, validation, Europe, health policy