UK Government consults on copyright and AI: Have your say!

Why should I read this?

On 17 December 2024, the UK government launched a pivotal consultation on copyright and artificial intelligence (AI). The initiative aims to address the evolving challenges and opportunities at the intersection of copyright law and AI technology. In this article we summarize the key points.

Background

The UK government faces the unenviable task of trying to find a position that supports both the creative industries and the AI sector. On the one hand, it wants to support rights holders’ control of their content and facilitate their remuneration where such content is used to train AI models, and on the other, it wants to support the development of world-leading AI models in the UK by ensuring wide and lawful access to high-quality data. The current lack of clarity regarding the copyright position benefits no-one: creative organisations are concerned their works are being used to train AI without their consent or any benefit to them, and AI companies are concerned about how they can legally access the training data they need, which risks stifling innovation. It is against this backdrop that the consultation seeks to gather input to establish the best way forward.

What are the key proposals?

Proposals regarding AI inputs

Data mining exception: In relation to the use of copyright material for AI training, the government proposes a new data mining exception. This would allow the use of copyright works, which are largely found by scraping data on the internet, for AI training without risk of infringement provided that:

• the AI developers have lawful access to such material
• the rights holders had not expressly reserved their rights by “opting out” using a machine-readable format.

There is a current exception under UK copyright law for data mining but this only applies to use for non-commercial purposes. This proposal would shift the burden from AI developers (to seek a licence in every case) to rights holders, who would need to opt out under this proposal. This position would bring the UK into closer alignment with the EU approach. The consultation asks whether this option is preferred to the alternatives of maintaining the status quo, requiring express licensing in all cases, or introducing a broader data mining exception which cannot be opted out of.

Technical standards: The consultation acknowledges that some tools to reserve rights already exist but have technical limitations, and so these would need to be further standardized and adopted to avoid rights holders having to deal with multiple systems.

Licensing: In addition to training AI models on materials available on the internet, some materials will be expressly licensed under contracts for this purpose. The consultation asks whether measures should be introduced to ensure that rights holders are not unfairly prejudiced, for example by blanket licence terms being bundled in with other terms that they need to accept in order to use certain platforms, tools and services. The consultation also proposes the use of collective licensing and aggregation/brokering services in providing access to works and remuneration for rights holders and asks for views on this.

Transparency: Central to the proposed solutions is the concept of transparency about the material used to train models and how this has been obtained. The consultation argues that greater transparency would allow rights holders to enforce their rights and increase trust in the AI development process. Measures to achieve greater transparency could include a requirement for AI developers to disclose the sources of their training materials, keep records, provide information on request or evidence compliance with rights reservations. The consultation includes questions around these to determine which approach would be most proportionate.

Wider clarification of copyright law: If a new data mining exemption is introduced, the government indicates that it would want to remove the following areas of ambiguity in the existing law.

• Treatment of AI models trained in other jurisdictions: At present these would be subject to the laws of the country in which training took place even if used in the UK, and so the consultation asks how developers can be incentivised to train their models here and comply with UK law.
• Temporary copies: There has been discussion around whether the existing temporary copies exception (which permits temporary copies to be made during technological processes) would apply to the training of AI models. The consultation asks whether clarification on this point is needed.

Encouraging research and innovation: The consultation asks how best to tailor the rules depending on the purpose of an AI model or size of AI firm to be proportionate.

Proposals regarding AI outputs

Protection available for AI outputs: The protection available for computer-generated works can arise under different sections of the Copyright, Designs and Patents Act 1988 including as a computer-generated work under section 9(3). Section 9(3) only applies to literary, dramatic, musical or artistic work which are “original”, the test for which is associated with intrinsically human qualities, making it unclear how a work could be both computer-generated and original. Many other countries such as the US and EU do not include protection specifically for computer-generated works and the provision was drafted long before AI came into existence. As such, the consultation proposes to remove it unless the sufficient evidence is received of positive effects of this protection.

Liability for AI-generated outputs: If these reproduce a substantial part of a copyright work then that will constitute copyright infringement. In this event, both the user and the provider of the AI model may be liable. The government indicates that it considers the existing copyright framework in relation to be adequate in this respect so no changes are proposed, but ask for views on whether the current approach allow for effective enforcement of copyright and what steps (such as keyword filtering) AI providers should take in order to avoid creating infringing outputs.

Output labelling: The consultation asks whether AI generated outputs should be labelled as such, as will be required in the EU for example, whilst acknowledging that such a requirement would come with practical difficulties.

Digital replicas (deepfakes): The consultation asks for views on whether the current laws protecting personality rights in the UK (i.e. passing off, copyright and performers rights) are fit for purpose or if more is required to combat deepfakes.

How to Respond

The consultation is open until 25 February 2025 and responses can be submitted online through the consultation portal. Evidence and examples should be included, as this will help the government understand the practical implications of the proposals.

Conclusion

The consultation on copyright and AI represents a significant step towards creating a legal framework for the development and use of AI, and if the proposals are followed, would bring the UK regime into closer alignment with the EU. It acknowledges that waiting for case law to bring certainty in this area would take too long and indicates that legislation is likely to be needed. AI developers, content creators and any organisations that make content and data publicly available on the internet should ensure they understand how the proposals may impact them and consider participating in the consultation to ensure that their interests are represented.