UK government opts out of its (previously) preferred proposal on copyright and AI

Last week, the UK government published its report on copyright and AI (and accompanying economic impact assessment) due under the Data Use and Access Act 2025 (DUA Act). Whilst largely deferring policy decisions until a later date, the report confirms that introducing a broad text and data mining (TDM) exception, which would have allowed use of copyright works for training generative AI (GenAI) absent a rightsholder opt-out, is no longer the government’s preferred way forward. The government also reiterated its preference to remove existing provisions protecting computer generated works (CGWs) in the absence of evidence of their ongoing value.

We take a closer look below.

Background

In December 2024, the UK government launched a consultation on AI and copyright (see our blog) (Consultation). That Consultation was relatively far reaching, but, perhaps most importantly, it contained the government’s proposals on how best to balance the rights of AI developers and IP rightsholders when training GenAI. Of the four policy options put forward, the government's stated preference was to introduce a broad TDM exception, which applies to TDM for any purpose, subject to rights holders having the ability to reserve their rights and opt their content out, underpinned by supporting measures on transparency.

That stated preference sparked a backlash from those in the creative industries, which, in turn, resulted in Parliamentary “ping pong” during the passage of the DUA Act because of the fierce debate around copyright and AI. A compromise was ultimately reached between the two Houses, with the government agreeing to publish, by 18 March 2026: (i) an economic impact assessment of the four policy options put forward in the Consultation; and (ii) a more general report on the use of copyright works in the development of AI systems (Report). It is these that were published last week. A summary of some of the key takeaways is set out below.

Key takeaways 

1. A broad TDM exception with opt-out is no longer the government’s preferred option. 

   This is not a surprise given previous acknowledgements from the government that it had made an error in noting its preference, and the nature of responses received to the Consultation (with 88% favouring strengthening copyright by requiring licences in all cases and only 3% supporting the government’s preferred option, as originally flagged in the progress statement issued in December). 

   The government has not put forward an alternative preference at this stage, opting instead to gather further evidence. However, it did note some additional approaches for further consideration (on top of the original proposals), including: (i) a “focused” exception, which could apply to specified use cases, such as science and research, or those in the public interest; and (ii) a similarly broad exception, accompanied by a statutory licence or levy (similar to that being explored by the Indian government).

   For now, that means that the status quo will continue, with the UK only having a very narrow TDM exception for non-commercial research purposes.

2. There is broad consensus favouring greater transparency over training data sources. 

   As things stand, there is no requirement in the UK for AI developers to publish details of the works used to train their GenAI models or how they were obtained. Responses to the Consultation indicated that there is broad consensus for greater transparency, but, unsurprisingly, there are differing views on approach - particularly on the appropriate level of granularity and how information should be disclosed. 

   In order to inform any future potential reform, the government will continue monitoring the approach taken in other jurisdictions, such as the EU (see our blog) and California, and work with industry and experts to develop best practice on transparency.

3. There are no immediate proposals to amend UK copyright law in respect of models trained abroad.

   The Report considered whether UK copyright law should apply to models that have been trained overseas and then placed on the UK market. A related question was recently considered in the Getty Images v Stability AI case, where the English High Court was asked whether importing a trained generative AI model into the UK could amount to secondary infringement of copyright (more on this here and here).

   The government noted the potential negative effect such extra-territorial reach could have, such as models not being made available in the UK, and the possible negative impacts on downstream developers of AI systems (which are more likely to be UK based than providers of the major GenAI models themselves). It concluded that this is an area of the law that should be considered properly by the English courts, based on the law as it stands. It also acknowledged that developments in other countries (particularly the EU and US) could have a significant impact on outcomes in the UK, with greater clarity expected in those jurisdictions in the next year or so. 

   On this basis, the government proposed not to amend UK copyright law in respect of models trained abroad for now, but will continue to monitor the situation and engage with other countries on the point.

4. The government will not intervene in the licensing market at this stage. 

   The Report also looked at the licensing of works for AI-training purposes, concluding that the government should not intervene in the market at this stage. Instead, the government proposes to continue monitoring both the licensing market, as it continues to develop, and other global developments and judicial outcomes that could impact it.

5. The Government’s preference to remove existing provisions providing copyright protection for computer generated works (CGWs) remains.

   The UK is one of the few countries in the world that provides copyright protection for CGWs, but there are questions around how those rules currently apply in a GenAI context. Acknowledging those issues, the government consulted on a number of potential reforms, with its preference being to remove the CGW provisions unless sufficient evidence emerged on the positive effects of protecting CGWs. The Report indicates that most respondents to the Consultation supported the removal of these provisions and the government has so far received limited evidence that the existing protections are being used. Whilst the government didn’t take the opportunity to make a final decision on this point, preferring to continue monitoring the use and impact of these provisions, the preference is still very much to remove these provisions unless strong evidence to the contrary arises. The government did, however, note that, if these provisions were to be removed, works created with the assistance of AI will continue to be eligible for copyright protection as for other original works.

6. Options to address risks relating to digital replicas (aka deepfakes) to be explored.

   Unauthorised digital replicas, including those mimicking a person’s image or voice, are an area of growing concern. As things stand, however, there are no specific personality or image rights in the UK and, whilst reliance on certain broader protections may be possible, the scope for such reliance is limited. The Report noted that there was clear support from respondents to the Consultation for enhanced protections for image and voice, but no clear view on what form such protections might take or who may benefit from them. The government has therefore proposed to explore potential options to address this, including by considering whether it may be appropriate to introduce a new personality right.

Comment

Whilst these publications provide some useful insight into responses to the Consultation and the potential impacts of some of the policy options put forward, the majority of the core policy questions remain unanswered. Disappointingly for both sides of the debate, this means that the uncertainty surrounding the interaction between copyright and AI in the UK is set to continue, with any potential legislative reform deferred to a later date.

Rightsholders will, however, be pleased to see the government confirm that a broader TDM exception with opt out is no longer its preferred option (although the government has not followed the recent Communications and Digital Committee of the UK House of Lords’ recommendation to clearly rule it out altogether (see our blog)). It will be interesting to see whether this retreat has any impact on the developing licensing market.

For now though, it looks like we’re back in a holding pattern, with little guidance on when we might expect to see any firm decisions being made. This prolonged uncertainty is not helpful for either side of the debate and, ultimately, could be bad for business in both the creative and technology sectors.