In a decisive response to mounting evidence of harm, the mental health charity Mind has initiated a pioneering global inquiry into the role of artificial intelligence in mental health support. The year-long commission, the first of its kind worldwide, comes directly after a Guardian investigation exposed how Google’s AI Overviews were providing “very dangerous” and inaccurate medical advice to the public.
The investigation found that AI Overviews, which use generative AI to create summaries atop traditional search results and are seen by 2 billion people a month, served up false and misleading information across a range of health issues. This included dangerously incorrect guidance on mental health conditions such as psychosis and eating disorders, which experts stated could deter people from seeking help or even put lives at risk. Specific examples uncovered included AI wrongly advising pancreatic cancer patients to avoid high-fat foods—the opposite of standard care—and offering oversimplified, context-free interpretations of liver function test results.
An Illusion of Definitiveness
Rosie Weatherley, information content manager at Mind, noted that while searching for mental health information online “wasn’t perfect” before AI Overviews, it typically directed users to credible websites offering nuance and lived experience. She argued that AI Overviews replace this with a clinical-sounding summary that creates “an illusion of definitiveness,” swapping contextual clarity for brevity in a way that is “seductive” but irresponsible. The Guardian also reported that Google was downplaying safety warnings about the potential inaccuracy of its AI-generated medical advice, and that repeating the same search could yield different AI summaries at different times.
Following the reporting, Google removed AI Overviews for some, but not all, medical searches. A Google spokesperson stated that the company invests significantly in the quality of AI Overviews, particularly for health topics, and that the vast majority provide accurate information, linking to reputable sources. They added that for queries indicating distress, systems work to display relevant crisis hotlines. However, Dr Sarah Hughes, chief executive of Mind, countered that “dangerously incorrect” mental health advice is still being served, which could reinforce stigma or prevent treatment-seeking.
The Double-Edged Sword of AI Support
The inquiry launched by Mind, which operates in England and Wales, aims to scrutinise this complex landscape. It recognises AI’s enormous potential to improve lives by widening access to support, offering personalised care through tools like chatbots, aiding early detection, and delivering therapeutic techniques such as cognitive behavioural therapy. Yet, it will equally examine emerging risks that extend beyond inaccurate summaries. These include users forming unhealthy dependencies or psychological attachments to AI, potentially leading to social withdrawal, and the provision of harmful advice on sensitive topics like self-harm or suicide even with safeguards in place.
Further risks involve the dissemination of misinformation without vital nuance, significant privacy concerns over sensitive mental health data, and algorithmic bias if training data lacks diversity, potentially leading to culturally insensitive or harmful responses. Regulatory bodies like the World Health Organization are already outlining considerations for AI in health, emphasising transparency, data quality, and clear risk management.
Mind’s commission will bring together leading doctors, mental health professionals, people with lived experience, health providers, policymakers, and tech companies to gather evidence on the intersection of AI and mental health. Dr Hughes stated the goal is to shape a safer digital ecosystem with strong regulation and safeguards, ensuring innovation does not come at the expense of wellbeing and that lived experience is central to shaping digital support. The charity, founded in 1946 and known for its campaigning work, has a board where at least half the trustees have direct experience of mental health problems, underscoring its commitment to this principle.
“People deserve information that is safe, accurate and grounded in evidence, not untested technology presented with a veneer of confidence,” Hughes said. The commission promises an “open space” where the experiences of those with mental health conditions are seen and understood, aiming to ensure the promise of AI is realised only through responsible development and deployment proportionate to the risks.
