In case you missed it before the holidays: on 17 December 2024, the UK Government published a consultation on “Copyright and Artificial Intelligence” in which it examines proposals to change the UK’s copyright framework in light of the growth of the artificial intelligence (“AI”) sector.
The Government sets out the following core objectives for a new copyright and AI framework:
- Support right holders’ control of their content and, specifically, their ability to be remunerated when AI developers use that content, such as via licensing regimes;
- Support the development of world-leading AI models in the UK, including by facilitating AI developers’ ability to access and use large volumes of online content to train their models; and
- Promote greater trust between the creative and AI sectors (and among consumers) by introducing transparency requirements on AI developers about the works they are using to train AI models, and potentially requiring AI-generated outputs to be labelled.
In this post, we consider some of the most noteworthy aspects of the Government’s proposal.
- The proposed regime would include a new text and data mining (TDM) exception
First and foremost, the Government is contemplating the introduction of a new TDM exception that would apply to TDM conducted for any purpose, including commercial purposes. The Government does not set out how it would define TDM, but refers to data mining as “the use of automated techniques to analyse large amounts of information (for AI training or other purposes)”. This new exception would apply where:
- the user of the work (e.g., the AI developer) has “lawful access” to the relevant works – including “works that have been made available on the internet, and those made available under contractual terms, such as via a subscription”; and
- the right holder has not expressly reserved their rights in relation to the work using machine-readable formats “which should be standardised as far as possible”. (If the rights have been reserved, a licence would be required to undertake TDM).
The proposal is similar to the EU Digital Single Market Copyright Directive’s TDM exception. That said, the Government acknowledges that there are limitations to relying on the EU’s approach as a precedent, for example because it is not always clear what constitutes a valid rights reservation. The Government acknowledges, for example, that while the robots.txt standard is widely used as a rights reservation protocol, it “cannot provide the granular control over the use of works that many right holders seek”. To avoid some of those perceived shortfalls in the UK, the Government wants rights reservations to be made using effective, accessible, and standardised machine-readable formats. It plans to engage with industry initiatives and international partners to achieve this goal.
- To underpin the TDM exception, AI developers may become subject to new transparency obligations in the UK
The Government makes it clear that it considers that any new TDM exception will only be effective if “underpinned by robust measures to ensure [AI] developers are transparent about the works their models are trained on”. This is on the basis that it is “difficult for right holders to determine whether their works are being used to train AI models”. As such, “increased transparency by AI developers will be crucial to ensuring copyright law is complied with and can be enforced”. The Government also considers that requiring transparency will improve the ability of AI developers, deployers of AI models and end-users to better assess their legal liabilities, which could encourage uptake of these technologies.
To achieve its aims, the Government states that regulation may be needed, and that it will “consider the case for it as it further develops its approach to AI regulation”. These transparency measures could include:
- Requiring AI developers to disclose their use of specific works and datasets;
- Disclosing ownership of web crawlers and the purposes for which content is being crawled; and/or
- Obligations to maintain records, to provide information on request, or to evidence compliance with rights reservations.
The Government recognizes that such transparency measures could be practically challenging for AI developers when “such a large quantity of works is used in the training process”. It therefore intends to consider whether (and if so, how) it should support the development of new tools to facilitate these transparency measures. It also acknowledges that transparency could be legitimately curtailed to avoid disclosure of trade secrets or information provided under a commercial contract.
Pointing to transparency obligations under the EU’s AI Act and California’s Assembly Bill 2013, the Government notes that it intends to engage with international partners and will consider making its measures interoperable with other international regimes. Under Article 53(1)(d) of the EU AI Act, for example, providers of general-purpose AI models must “draw up and make publicly available a sufficiently detailed summary about the content used for training of the general-purpose AI model”, using a template provided by the EU’s AI Office. For more information, please see our blogpost on the obligations under the AI Act, available here.
- The new regime will clarify the status afforded to AI-generated content
The Government is proposing to reform the copyright protection of “computer-generated works” (CGWs), a somewhat unique form of copyright available in the UK and a small number of other jurisdictions (e.g., Ireland, India, and Singapore) that affords protection to works “generated by computer in circumstances such that there is no human author of the work”. The UK Government has previously consulted on whether it should maintain this protection in UK copyright law, but the results of that prior consultation were inconclusive. In this latest consultation, the Government discusses certain critiques of maintaining the status quo, among them that the originality requirement under UK copyright law applies to works that are an “author’s own intellectual creation”, something that cannot necessarily be replicated by a machine. Discussing various policy options available to it, the Government notes that it might consider either amending the existing protection (e.g., removing the requirement that a CGW must be “original” to benefit from protection, or clarifying when a CGW should be considered “original”), or removing the protection offered to CGWs altogether.
Under the latter option, for which the Government expresses a preference, they state that works that are “AI-assisted but which exhibit human creativity would continue to be protected. AI-generated music and video could continue to be protected as sound recordings, films, and broadcasts”. As such, “the main effect of removing protection (if any) would appear to be on content generated by AI without a human creator which takes the form of images or text”.
- AI developers could be required to label outputs as AI-generated
The Government consultation also assesses whether changes to copyright might be needed in relation to AI outputs (i.e., works generated by AI). One area where it is considering introducing legislation relates to the labelling of AI-generated outputs. Any regulatory intervention could be based on the requirements under the EU AI Act, which imposes transparency rules for content produced by generative AI. On the other hand, the Government is aware of the practical issues associated with these labelling requirements. For example, it acknowledges that it’s not immediately clear whether a standard image that has been retouched using AI should be labelled as AI-generated, and that making labels resistant to manipulation could prove difficult. Currently, the Government has committed only to “consider whether and how it could support research and development into labelling tools that support transparency that is valuable for copyright”.
- Further changes to UK copyright law could be made to account for developments in AI
Were the Government to introduce the new TDM exception, it will also “ensure that existing ambiguity in the law is removed”. For example, the Government proposes clarifying how the “temporary copies” exception applies to the training of generative AI models; ensuring that the UK’s copyright regime is “internationally interoperable” (although the consultation paper does not explain what this means in practice); and potentially introducing measures “to establish a level playing field between providers of models which are trained within the UK, and those trained outside the UK but made available for use in the UK market”.
- Legislation will probably be needed
The Government’s consultation remains open to feedback until 25 February 2025. The Government states its intention to design the best possible policy approach after it has considered stakeholder responses. Once the Government settles on its final approach, it expects that “legislation is ultimately likely to be needed” to bring this new framework into effect – particularly given its view that ongoing litigation in this area, such as the Getty Images v Stability AI case in the UK, may not provide sufficient legal certainty in a timely fashion. That said, the Government does not entirely rule out bringing these changes into effect through other means.
****************
Covington’s UK technology regulatory practice is tracking these developments closely. Please feel free to reach out to your usual Covington contact, or the authors of this post, should you have any questions about the impacts of this proposed regime.