Let's talk AI: Understanding public attitudes towards use of AI with sensitive data.

Abstract

The development and use of Artificial Intelligence (AI) with sensitive data in Trusted Research Environments (TREs) is a complex and rapidly evolving area of work. To ensure the use of AI with sensitive data meets public expectations, a deeper understanding of public attitudes is essential. In late 2025, as part of the UK Research and Innovation DARE UK (Data and Analytics Research Environments UK) programme, we conducted a pilot public deliberation with 22 individuals in England with limited prior knowledge of data research and AI. A series of three workshops with the same group tested approaches to knowledge development for this emerging and complex topic, and explore hopes and concerns about AI. Discussions explored training and development of AI models in TREs, and use of AI capabilities to support data analysis and risk assessment. Participants recognised the potential of AI to improve efficiency and accuracy when drawing conclusions from data, delivering public benefit. While existing TRE safeguards reassured participants about personal privacy, fears of data and AI misuse persisted. Trust, transparency, oversight and regulation emerged as critical themes, and participants emphasised that AI should remain a tool, not a decision-maker. Participants called for diverse and representative AI training data to mitigate bias, and greater efforts to increase public awareness. These insights will inform a larger, UK-wide public deliberation on the same topic in 2026, in partnership with ADR UK (Administrative Data Research UK) and others. We look forward to sharing the interim findings of this wider work in July 2026.