UK researchers training AI on copyrighted articles or personal health data currently operate in a legal grey zone where they cannot easily share datasets with colleagues or collaborate with industry partners. The problem is that the UK’s 1988 copyright law only permits text and data mining for non-commercial research, and the Data Protection Act adds further restrictions on using personal data. This means universities cannot pool data to build large, reliable AI models, and the insights those models produce risk being legally questionable or scientifically weak. The project will identify exactly which legal hurdles cause the most trouble and propose a new framework that exempts all pre-market research from these restrictions, with transparency rules kicking in only when a product reaches the public. If the proposed reforms succeed, researchers could freely share and retain curated datasets across institutions and sectors, enabling larger-scale AI training for everything from drug discovery to climate modelling. The UK would also regain a competitive edge in AI-driven science by removing the current risk-averse fragmentation of data assets.
View original technical description
Recent advancements in artificial intelligence (AI) have catalysed transformative reform in scientific research. These AI-driven methods rely on vast quantities of information, including copyright-protected works (e.g., scholarly articles, software code, and literary texts) and personal data (e.g., health records, social media activity) to train and validate models. Yet current UK legal frameworks, specifically the text and data mining (TDM) exception under the Copyright, Designs and Patents Act 1988 (CDPA) and UK Data Protection Act, introduce significant legal uncertainties and restriction for researchers, such as who can use text/data from where for what purpose. While the existing copyright TDM exception can be interpreted as permitting certain AI training activities, it only covers non-commercial research and does not extend to the whole lifecycle of AI models and systems; together with data protection regimes, it creates ambiguities at virtually every stage of AI-based research and development. The existing copyright exception’s narrow scope, focused on non-commercial research, disallowing wider data sharing, and ambiguities on lawful access, does not reflect modern research practices. For instance, universities cannot easily collaborate with each other or with industry partners to create, retain, and share newly curated datasets, leading to major constraints on the breadth, scale, and reliability of data available for advanced AI research. Such restrictions, related to whether and when personal data can be used in AI-driven research, have been amplified by the UK Data Protection Act. Many data protection measures, such as consent, right to inform data subjects, have become hurdles for researchers seeking to use personal data in AI research. This risk aversion fragments available datasets, restricting both interdisciplinary research and the creation of high-value data assets for AI research. As a result, AI-generated insights risk being legally tenuous, underutilised, or epistemically unreliable. Without clear legal frameworks, researchers are fettered, and the potential of ‘AI for science’ cannot be fully unleashed. This project aims to identify the most important uncertainties and their effects on researcher and remove unnecessary legal hurdles by transforming the regulatory paradigm of current copyright and data protection regimes across the whole lifecycle of AI in research. It will propose policy and governance solutions that the UK should embrace a broader TDM and data protection exemptions to cover all research conducted prior to market entry, at which point transparency and licensing obligations would take effect. This could ensure responsible data practices and foster a vibrant environment for AI innovation in science. The project will confirm/reject these nuanced hypotheses by: Empirically investigating how researchers and institutions negotiate complex copyright and data protection constraints. Assessing the adequacy and implications of the existing copyright TDM exception and data protection legislation to determine how and why they fall short of supporting modern AI-driven research. Proposing a forward-looking regulation framework to align the needs of AI-driven research with updated legal exceptions and regulatory policy and guidelines. Potential applications and benefits include clearer legal guidelines and frameworks, enabling greater public-private research collaboration, improved data access and retention practices, and enhanced epistemic reliability of AI-powered research. The project will produce actionable policy recommendations for institutions (e.g., UKRI), the UK government, and regulators (e.g., IPO, ICO), fostering an open, innovative, and trustworthy environment for the responsible use of AI for science, thereby re-establishing the UK’s global leadership in scientific advancement.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know