The UK government has published a progress report on its copyright and AI consultation, which ran earlier this year. The final report will be available on 18 March 2026.

The 50-question consultation sought views on accesssing and using copyrighted-protected works to train AI models. It also invited views on transparency measures,  enforcement, and copyright protection of AI-generated works. The consultation received over 11,500 responsers which, remarkably, were analysed manually by a team of around 80 humans – no AI was involved in the analysis.

Views on copyright and AI are polarised. The debate is often perceived – and this is reflected in the consultation itself – as a split of views and interests between the creative industries and AI developers. This picture is reinforced by high-profile court cases, including Getty images vs Stability AI, Bartz vs Anthropic, Authors Guild vs OpenAI, and Disney/Universal vs Midjourney. Questions – and diverse opinions – arise around what is lawful, what may be covered by a copyright exception (or, in the US, ‘fair use’), whether licensing should always be required and to what extent AI developers shoud be transparent aroubd their sources. Overall the creative industries strongly favour a licensing approach, claiming that using images, works of fiction, music and other works to train AI models must always require a licence. AI developers mostly favour broad exceptions, claiming that such uses are covered by fair use or similar principles.

The interim report broadly supports this picture:

• Responses were from across all sectors. There were responses from the creative industries, the tech industries (including AI developers), academic and research organisations and cultural heritage organisations – although exact representation from different sectors is yet to be published. In addition to institutional responses there were also many individual ones.
• The overwhelming majority of the responses (88%) supported licensing in all cases. Unsurprisingly, it seems that most of those supporting a licensing approach in all cases are from the creative industries.
• On the other end of this debate lies a preference for a broad exception – similar to ones applied in Japan and Singapore – under which use of copyright-protected works to train AI may be permitted. Only 0.5% of respondents supported this option. Further, only 3% of respondents supported the government’s preferred approach: a broad exception allowing use of copyright-protected works for AI training unless the copyright owner has explicitly reserved their rights, opting out of the exception. This approach is similar to existing EU exceptions. AI developers appeared to support one of these options.
• Importantly, the report mentions that many respondents put forward proposals for new or modified exceptions to support activities such as research.

Implications for research and education

At this stage, it is not clear to what extent academic-research sector has contributed to the consultation; nor have the concerns and interests of research and education sectors been voiced as broadly as those in the creative industries. What is certain is that our sector needs clarity to continue using AI (which is not always generative AI) in education and research.

The current UK exception permitting text and data mining of lawfully accessed materials for non-commercial research purposes can only support research activities to some extent: in practice uncertainties and barriers often arise around the use of AI, lawful access and the sharing of copies among project partners. While the exception would benefit from further clarity, there is also a case for expanding it to ensure that AI can be used in broader contexts, including educational uses and collaborations with commercial partners. There are also concerns that licensing and rights reservation would add to the existing barriers, leading to re-licensing resources that research institutions already pay to have access to.

In this debate, the views and concerns of UCL researchers are important. The Copyright and TDM workshop we ran in November was an opportunity for UCL researchers to discuss examples where use of resources for computational analysis is essential. We are following up with a Copyright and AI session on the 4th of March. We will share further details and a registration link in January.