UK plans to be an "AI maker" not an "AI taker" with its new copyright consultation and AI Opportunities Action Plan

Diving deeper into the consultation paper and AI Plan

Consultation paper

Copyright, training data and a TDM regime

The consultation paper examines the topic of AI models, copyright works and training in detail. It highlights the challenges for rights owners in determining if their works have been used for AI model training. The consultation acknowledges that creators find it difficult to control and be remunerated for the use of their works, but that the AI sector needs access to a broad range of high-quality data.  However, questions remain over the legality of their activities.

The government's position is that the existing legal framework is not acceptable, which impacts both the creative and AI industries. It is in favour of regulation as a means of resolving current uncertainty and explores a range of policy options. While different options are tabled, the focus is on the introduction of a new broader TDM exception, a rights reservation model, transparency measures and proposals for standardising the approach used by rights holders and AI developers in order to be effective and facilitate compliance.

We expand on each aspect below.

Broader TDM exception

One proposal is that AI developers would be permitted to train AI models using copyright works for any purpose – including for commercial purposes - as long as they have lawful access and rights owners have not reserved their rights.

The new TDM exception is a significant expansion on what currently exists. The current UK TDM exception found in section 29 of the Copyright, Designs and Patents Act 1988 ("CDPA") only applies to research for non-commercial purposes. The new TDM exception is also similar to that used in the EU Digital Copyright Directive, though the UK proposal is wider in application.

If a TDM exception is progressed, the UK government is also interested in reviewing how the "temporary copies" exception operates and whether the UK TDM's existing exception for non-commercial research remains appropriate.

Rights reservation

There is a proposal which would allow rights holders to expressly reserve their rights to prevent use of their works for AI training. The government's proposal is that rights should be reserved using effective and accessible machine-readable formats (which should be standardised as much as possible) to assist with the proposed TDM exception. This is consistent with the approach in the EU.

The experience of the EU shows that it is not always clear what constitutes a valid rights reservation and that there are challenges brought on by a lack of standardisation and / or widespread adoption of different protocols and processes.

Given the experience of the EU, the UK government seeks submissions on ways that effective rights reservation can be achieved. It also emphasises that any future regulation must focus on achieving desired outcomes without being overly prescriptive, to ensure we can adapt to technical advancements.

Transparency

The government supports increased transparency from AI developers to improve understanding of how content is used, assess legal responsibilities, and promote the adoption and deployment of new systems. This enhanced clarity will help ensure compliance with copyright law and facilitate its enforcement.

Transparency measures considered include AI firms:

• disclosing use of specific works and datasets used in training data;
• summarising details of web crawlers (including ownership and the purposes for which content is being crawled);
• requirements to keep records and to provide certain information on request (including evidence as to compliance with rights reservations opt-outs). 

Recognising the practical challenges that AI developers may face with disclosure, the paper specifies that reporting need not be extensive and might be in summary form – similar to article 53 of the European Union Artificial Intelligence Act ("EU AI Act") or the California Assembly Bill 2013.

The transparency measures proposed by the UK are another way in which the interests of copyright owners are being respected. Transparency has long been a concern for rights holders so implementation will be welcomed from that quarter. Greater transparency also correlates with the position of the EU and some of the early regulatory developments in the US. On the flip side, transparency will be a concern for AI developers who may see this as a path to further litigation as rightsowners gain better visibility of activities. It also remains to be seen whether a summary will give owners the insight they are looking for. This may become clearer when we see how effective a similar initiative turns out to be under the EU AI Act over the coming year.

Standardisation

The use of standardised technical measures is suggested to address some of the experiences and challenges faced in the EU in relation to the recognition of rights holders' reservations and the application of measures by AI developers. For example, introducing standardised rights reservation protocols or processes similar to robots.txt standard (which can block the main generative AI web-crawlers). AI developers generally respect those that exist already. Another approach mooted is to associate metadata with the work itself which would flag to AI models and developers that it can or cannot be used. The UK government confirms that it is ready to engage with these standards initiatives being taken forward by industry and international partners, such as the EU, to ensure a collaborative approach. However, questions lie ahead in terms of the finer details and how government aspirations can be practically implemented.

Notably, the UK government recognises that there may be a need for it to encourage widespread adoption and compliance of protocols through dispute resolution mechanisms and oversight achieved in alignment with, or even outside of, the existing copyright framework. This is an interesting proposition which we will monitor.

Computer generated works ("CGWs")

The generative AI boom means that AI now creates outputs which are on par with human-made works. This has spurred the UK government to reconsider copyright protection for AI-only generated content as CGWs (which are literary, dramatic, musical, or artistic works which are purely computer generated with no human author).

The government questions the effectiveness of CGW protection currently provided under the CDPA. The "originality" requirement for literary, dramatic, musical, and artistic works involves the "author's own intellectual creation," which AI-created works often lack. There is also uncertainty about who uses the protection afforded to CGWs and whether protection is in fact needed.

The government is open to modifying or removing CGW protection. This might be by reforming the CGW provision to clarify its scope (by clarifying how originality works in the context of CGWs or by removing originality requirements). Alternatively, it might be by removing protection for CGWs altogether which marks a departure from the previous decision to leave the law as it is. Thismay lead to more alignment with other countries which do not provide CGW protection.

No particular issues were detected with the way the current infringement provisions apply, which the government deemed reasonably clear and adequate. However, questions remain over whether the current approach to liability in AI-generated outputs allows effective enforcement of copyright and welcomes views on specific areas where copyright law is considered to be deficient.

Separately, the government indicates interest in ensuring that AI generated outputs are labelled as being AI-generated, but notes the technical challenges of introducing labelling. It invites parties to propose technical methods to enable labelling of AI outputs, ensuring that they are resilient to manipulation.

Licensing

The consultation highlights the important role that licensing has in ensuring that rights holders are compensated, while AI developers (particularly smaller entrants) gain access to quality training materials. The paper encourages collaboration between the creative and AI industries to develop methods that improve IP licensing. The government seeks information on how it can foster effective licensing practices and whether these practices enable rights holders to maintain control.

Other emerging issues – digital replicas, AI in education, AI inference and synthetic data

The consultation also raises a series of emerging issues which it invites feedback on.

It identifies concerns involving the use of digital replicas (better known as "deepfakes"). To manage these concerns from a copyright perspective, the paper considers the possibility of individuals reserving their rights at the input stage of AI training which might otherwise use digital replicas of individuals' voice and appearance. Although the UK is not actively developing a formal policy on personality rights at this time, it invites feedback on a more comprehensive personality right from interested parties. However, the government recognises that the introduction of a new type of IP protection for personality in the UK would be a significant step and requires careful consideration.

It also invites views on the use of AI in education and how AI products interact with copyright works by inference (ie the process by which a trained AI system generates outputs using new data) and the increasing use of synthetic data to train AI models.

AI Plan

Copyright initiatives have found a place amongst the 50 recommendations set out in the AI Plan. The following will be of interest to rights holders and the AI industry:

Creation of a new National Data Library and initiatives for higher quality datasets

Recommendations 7 to 13 of the AI Plan table a range of dataset initiatives which are aimed at identifying and unlocking value so that greater AI development can occur. Briefly, these recommendations include:

• identifying high impact public datasets for use in the National Data Library which will be made available to AI researchers and innovators. The government has responded to this recommendation advising that the Department for Science, Innovation and Technology ("DSIT") will explore how the initiative can be executed with further details expected in June 2025;
• incentivising industry and researchers to develop and open datasets; and
• setting up a "copyright cleared" British media asset training dataset which is licensable globally. This dataset could potentially include material from the BBC (its extensive media archives), the British Film Institute (its vast collection of British films and media) and National Archives (its historical documents and media). The government intends for the DSIT and the Department for Culture, Media and Sport to discuss the government's role in this initiative with partner organisations and industry from March 2025 onwards.

The introduction of a copyright cleared media asset dataset is an interesting proposition which we will continue to monitor. Several questions spring to mind around this initiative. For example, how will this dataset impact industry and could it establish a precedent for forward looking datasets? How will copyright clearance and licensing mechanisms be executed and monitored for compliance? And of course, how does this interact with the wider litigation that is currently on foot over the use of copyright works in training datasets?

Regulatory reform and AI champions

The AI Plan, in a similar but more explicit fashion than the consultation paper, refers to the UK's current pro-innovation approach to regulation which the AI Plan holds up as an asset for the UK vis-à-vis other jurisdictions. 

Building on that, the AI Plan presses for a new TDM regime which "is at least as competitive as the EU" as part of recommendation 24.

The plan also calls for consideration of the use of AI sector champions in key sectors like the creative, life sciences and financial services industries with champions working with industry and the government to develop AI adoption plans and promote AI innovation. This is a novel proposition and one that could have an important strategic bearing for those industry leaders who find themselves as champions if this recommendation is progressed.