Support for sharing health data for AI hinges on safeguards, consent and public benefit

Public support for sharing health data for artificial intelligence research depends on clear public benefit, strong safeguards, and meaningful consent, according to a new study based on in-depth focus groups with members of the U.K. public. The findings add to a growing body of evidence about public views on health data sharing for AI research, as governments are also promoting data sovereignty and tightening control over sensitive biomedical datasets.

The researchers conducted eight online focus groups with 41 adults from across the U.K., selected to reflect a range of ages, ethnicities, health experiences and socioeconomic backgrounds. Participants discussed realistic scenarios involving health data sharing for AI, including university-led research, large research databases, and projects involving commercial companies.

Across the discussions, participants expressed cautious and conditional support for health data sharing. Anonymization was widely seen as essential, but not foolproof, particularly for people with rare conditions or where large datasets are linked together. Many participants accepted that some level of risk was inevitable but wanted greater transparency about how data are protected and what would happen if things went wrong.

Trust varied depending on who was using the data. Universities and the NHS were generally seen as acting in the public interest, while the involvement of commercial organizations prompted greater skepticism. However, that view was lessened when commercial involvement could be clearly linked to patient benefit and subject to strict oversight.

Participants made decisions about whether to share data by weighing perceived risks against potential benefits to themselves and others. Concerns about discrimination, misuse and future unknown risks were weighed against potential benefits such as improved care, faster diagnosis and helping future patients. Many described concern for the well-being of others and the "greater good" as an important motivation, particularly those with long-term conditions or previous experience of benefiting from medical research.

Consent emerged as a central issue as a foundation for trust. Participants wanted information that was clear, specific and relevant to the particular study, and in an accessible format. They also emphasized the importance of the process of seeking consent, opposing requests made during stressful or emotionally vulnerable clinical moments. Suggestions included tailored approaches, opportunities to opt out of certain uses of data, "cooling-off" periods, and the ability to withdraw consent at a later stage.

The debate is unfolding as the geopolitics driving artificial intelligence superpowers is reshaping biomedical datasets, and who has access to them. After decades of policies pushing open science, governments are now promoting data sovereignty — the idea that sensitive datasets should remain under national control and foreign access should be conditional.

In Europe, nations are feeling pressure to keep the continent competitive in AI while staying inside strict privacy rules and reinforcing European data sovereignty. In November 2025, the European Commission launched RAISE, the Resource for AI Science in Europe, a virtual institute to coordinate AI resources, including scientific datasets, across member states. According to the Commission, databases such as Europe’s Genomic Data Infrastructure and Cancer Image Europe will be harnessed to serve planned public-private AI gigafactories, and Horizon Europe will channel €600 million to secure compute time for researchers.

Researchers have warned that as nations and companies raise walls around health records, algorithms will be harder to validate across different populations, more prone to hidden bias and less likely to benefit the people who contributed their data in the first place. In Europe, data are available for AI development, but only through controlled, application-based mechanisms.