The UK government has launched a new consultation on plans to reform copyright law to address some complex and contentious issues when it comes to the intersection of AI and IP law. 

Below, we summarise the key points. The consultation is open for comment until 25 February 2025. 

Expansion of the TDM exception

The process of AI training requires very large amounts of data, which are often protected by copyright. There is significant debate about whether data mining such works and using them for AI training without consent amounts to infringement, and about how to balance the competing interests of AI developers and rights holders.

The government is seeking comment on its plan to permit text and data mining (TDM) for commercial purposes (the current TDM exception in the Copyright, Designs and Patents Act 1988 (CDPA) only pertains to non-commercial research), underpinned by supporting measures on transparency. The TDM exception would apply by default to all works to which the user has lawful access, including all works freely available on the internet, as well as any the user has paid a subscription to access.

However, a rights holder would have the ability to "reserve its rights", allowing it to opt out of the exception and thereby not permit data mining of its materials. A licence to do so would then be required, allowing rights holders to seek remuneration. This approach would be further underpinned by measures requiring greater transparency from AI developers about the sources of training material. 

Such proposals would bring the UK's approach more closely in line with the EU regime under Article 4 of the Digital Single Market (DSM) Copyright Directive (Directive (EU) 2019/790).

The government has asked for input on whether respondents agree with its proposed approach, or whether other options are preferred, including doing nothing, strengthening the requirement for express licensing in all cases or introducing a broad TDM exception without the ability to opt out. 

The consultation also seeks views on how the reservation of rights and transparency provisions should work in practice, given the experience that the EU TDM exception regime to date has produced uncertainty about what constitutes a valid reservation of rights in practice. It therefore wishes to adopt technical measures that make it as clear as possible to all parties whether a rights holder has opted out of the TDM exception and therefore proposes that reservation of rights is done using standardised, effective and accessible machine-readable formats. 

Consultation is therefore sought on various technical standards for achieving this:

• Robots.txt – also called the Robots Exclusion Protocol, an exclusion protocol page which indicates to web crawlers and other TDM robots which portions of a website they are allowed to use.
• Metadata – including instructions on whether a work is available for training within the metadata associated with each work.
• Direct notification – notifying AI firms and dataset owners directly that they cannot use certain works used in training, which may also involve third party "do not train" registries. 

The consultation notes various shortcomings and difficulties with all the current standards, and believes further standardisation is needed. For example, the direct notification model is difficult to manage, whereas the widely adopted robots.txt standard cannot provide granular control. The UK government therefore welcomes detailed views on how to avoid overly prescriptive provisions while still achieving the desired outcomes above. 

Transparency provisions 

The UK government considers that robust transparency measures are key to building trust in the new approach and has said it will not introduce a TDM exception without them. The proposals seek to address concerns around AI developers not clearly disclosing the source of their training data, making it difficult to determine whether certain works have been used in training (allowing rights holders to enforce their rights) and creating uncertainties around assessing risks and liability for those implementing AI tools. 

However, the consultation recognises the practical challenges transparency measures could pose given the large quantities of data involved, as well as legitimate reasons to restrict such information (e.g. in the case of trade secrets). 

Input is therefore sought on how to implement transparency requirements in a balanced and proportionate way, including comments on the EU's approach requiring AI developers to publish a "sufficiently detailed summary" of training content. The EU regime does not require exhaustive reporting, as long as the main sources are cited and a template report is being developed. It therefore looks as though the UK will take the EU model as a starting point. 

Protection for the outputs of generative AI

The consultation is also seeking views on section 9(3) of the CDPA, which provides that computer-generated works without a human author are taken to be owned by the person who made "the arrangements necessary for the creation of the work to be undertaken". 

The UK government notes there are two main criticisms of the exception. The first is that it appears incompatible with the requirement that copyright works must be original in the sense that they express free and creative choices and reflect the "personality" of the author, all of which are human qualities that a work without a human author seemingly cannot possess. The second is that the provision is not needed. The UK government considers that works produced where a human uses an AI tool are already protected as AI-assisted works and there is little if any need or incentive to protect works that are pure AI creations. 

The UK government therefore proposes to do away with the exception entirely, but seeks input on whether it should do so or instead maintain the current provisions or alter their scope. 

The consultation also asks for input on various other areas of uncertainty around AI and copyright, including whether there should be mandatory labelling provisions for AI-generated works and whether current contractual and licensing practices give the original creator or performer sufficient control over the use of their works for AI training.

Conclusion

These proposals would be a significant development for the UK AI and creative industries. 

Largely mirroring the EU's approach will help minimise the impact for rights holders who are already adopting technical measures to reserve rights in response to the provisions in the DSM directive. The EU regime has its critics, but the UK government also appears to be taking lessons from the EU's implementation by looking to provide more certainty on how the measures will work in practice. Greater clarity on the mechanisms to reserve rights and opt out of the TDM exception will be of particular importance to ensure the success of the proposals. 

AI innovators, however, may consider that the inclusion of opt-out measures does not adequately balance the protection of copyright holders with the need for fostering innovation in AI and may lead to a "chilling effect" on AI research. 

The consultation will run for 10 weeks, closing on 25 February 2025. Those wishing to read more about the consultation and give feedback can do so here.