AI and Copyright: UK government backs away from exceptions for AI training, proposes maintaining the status quo – at least for now

Background

From December 2024 to February 2025, the UK government ran a consultation on copyright and AI, seeking input on a number of options that attempted to strike a balance between AI developers being able to use copyright works when training models, and the rights of copyright holders. The consultation also sought views on other areas, including copyright protection for computer-generated works, digital replicas and AI labelling.

Following the consultation, several copyright related obligations were put on the government in the Data (Use and Access) Act 2025, including an obligation on the government to publish a report, and impact assessment, on the use of copyright works in the development of AI models by 18 March 2026. In a written statement Liz Kendall (Secretary of State for Science, Innovation and Technology) said that the report fulfils the government's obligations under the DUA Act and sets out their next steps on copyright and AI.

Report on Copyright and AI

The government's overall approach is to maintain the status quo in relation to using copyright works to train AI models. It will not make any legislative reforms at this stage, while it continues its discussions with stakeholders, builds up more evidence on the impact of copyright on AI development in the UK, considers alternative proposals put forward in responses to the consultation (for example, a focused exception targeting certain types of use) and while it sees how things develop in other countries, such as the US. There are however several areas identified as areas for further work: digital replicas, labelling AI content, creator control and transparency and independent creatives. And the government has said it will remove the specific protection for computer generated works.

Copyright exception with opt-out

In the consultation, the government's favoured option was a text and data mining ("TDM") copyright exception, with a rightsholder opt-out. However, the report says that this is no longer the government's preferred option – according to Liz Kendall's statement, the government has heard the concerns of the creative industries and there is now no preferred option.

The report says the government wants to gather more evidence on the impact of copyright on the development and deployment of AI. It plans to discuss other potential approaches with stakeholders, such as a focused exception targeted at certain types of use or application, which it says was the most popular alternative proposal put forward in consultation responses. It will also continue to monitor developments in technology, litigation, international approaches and the licensing market. In particular, since a copyright exception with opt-out is similar to the EU approach, the government wants to see how EU case-law helps to clarify the scope and effect of the EU exception. The government does however rule out a broad exception with no opt-out.

Digital replicas

The government has identified digital replicas of people's voices or likeness (also known as deepfakes) as an area of harm that needs addressing. The consultation sought views on whether the current legal framework for protecting a performer's likeness remains fit for purpose, including whether individuals have sufficient control over their personality. In the report, the government says the responses to the consultation indicated that this area is an area of growing concern, particularly for performers, with respondents highlighting that copyright protects creative works but not the people featured in them. But there was no single view on what form enhanced protections for a performers' voice or likeness should take. The government has therefore decided it will explore a range of options including whether to introduce a specific ‘personality right' into UK copyright law and the government will launch a consultation in the summer seeking views on the options.

AI labelling

The government recognises that AI labelling can be helpful for consumers in understanding whether content has been made using AI, and also to help protect against deepfakes and misinformation. The report says that the responses to the consultation on the topic of AI labelling highlight the need for one approach to wholly AI generated content and a different, more nuanced, approach to AI-assisted content. The government proposes establishing a taskforce to put forward proposals for government on best practice for labelling AI-generated content, with an interim report to be published in autumn.

Creator control and transparency

The government understands that transparency from AI developers about the works they have used to train their models is fundamental to enabling rightsholders to assert their rights and also for AI developers to demonstrate compliance. However, there are differing views on the level of transparency obligations that AI developers should be put under, with AI developers preferring high level voluntary obligations but with rightsholders seeking mandatory granular detail. The government says it plans to work with industry and other experts to establish best practice. Liz Kendall's statement says the government will publish a review of the mechanisms available for creators to control their works online. This will include standards, technical solutions and best practice on input transparency. This review will inform where there are gaps and whether there is an appropriate role for government in addressing them.

Independent creatives

The government will launch a working group on independent and smaller creative organisations to explore whether there is a role for government to support their ability to license their content.

Comment

The government has published a detailed and thorough report and economic impact assessment but over a year after the copyright consultation ended, and months of discussions with working groups on the issues, the government has still put forward no concrete decisions or legislative proposals on the use of copyright works for training AI models. The only legislative change it has said it plans to make is to remove the long-standing existing protection for computer-generated works in the Copyright, Designs and Patents Act.

Rightsholders will be relieved that the government appears to be significantly rowing back on the introduction of a text and data mining exception with opt-out. It hasn't been ruled out completely (unlike a broad exception with no opt-out) but it seems that in practice this option is probably now dead, which would mean the UK diverging from the EU, with regard to TDM for commercial uses.

Although the government did not make any specific proposals in relation to combatting digital replicas in this report, the government has clearly identified this area as a high priority area for reform, and we expect that following the proposed consultation in the summer, this is the area that we will see the government put forward legislation in the fairly near future. Look out for our deeper dive on the law on digital replicas and developments in this area soon to come.

Finally, although the government has not taken on board the specific recommendations of the House of Lords Communications and Digital Committee, published in their report on 6 March 2026 (see our earlier article here), the government has said that they will carefully consider the recommendations and take them into account going forward. So at this stage, all we can say is that the government will maintain the status quo in relation to copyright and training AI models for now, while it continues its work in this area, but whether that will change remains to be seen. The debate is not yet over.

Authored by Penny Thornton and Alastair Shaw.