Creative Rights in AI Coalition—written evidence (AIC0001)

House of Lords Communications and Digital Select Committee inquiry: AI and copyright

The Creative Rights in AI Coalition

The Creative Rights in AI Coalition is a broad group of over 70 rights holders, AI businesses, and organisations from across the creative industries. We are calling on the Government to spur growth in the creative and tech sectors by protecting copyright. A full list of members can be found here.

Principles

The Coalition has developed three core principles which should guide the Government’s policy-making around AI and copyright.

IP and a dynamic licensing market

UK copyright law grants intellectual property owners, including tech companies, exclusive rights over their works, which includes control over the granting of a licence for uses of those works. Copyright protected works are a valuable resource for building and operationalising generative AI tools, products and services. Despite acknowledging this value, generative AI developers have largely exploited such content without permission, ignoring copyright protections and clear reservations of rights. A mutually beneficial, dynamic licensing market is feasible and desirable but can only flourish if there is respect for copyright, underwritten by robust mechanisms to ensure accountability and compliance.

Transparency

The Government should focus on solutions which ensure the creative industries can meaningfully exercise their exclusive rights, including if and how their copyrighted content can be collected and used by generative AI developers. Transparency is therefore essential in supporting accountability for copyright infringement and must be designed to incentivise compliance with copyright law, fostering a mutually beneficial partnership between generative AI developers and the creative industries.

Growth and innovation

The UK creative industries are an economic powerhouse with huge growth potential, as well as wielding immense cultural and soft power significance. A dynamic content licensing market, underpinned by robust copyright protections, will drive growth and innovation in the Creative Industries and generate value for society and the UK economy at large. This is also essential for future growth in the generative AI sector, which is reliant on a sustainable supply of high-quality, human-authored copyright works, without which it risks collapse.

Questions

1.              Are there uncertainties, gaps or barriers in the UK’s current copyright framework that restrict innovation or rightsholders’ ability to enforce their rights in relation to generative AI?

a)     If so, how could these be addressed? 

UK copyright law is absolutely clear: rightsholders must consent before their work is used by AI developers for commercial purposes. Unfortunately, the Government had generated uncertainty with a drawn-out consultation process and wrongly questioning the clarity provided by UK copyright law, which has undoubtedly disincentivised AI developers from striking deals with UK rightsholders. The Government’s top priority must be to rule out any new text-and-data-mining exception (or other weakening). This will be key in supercharging an already growing licensing market.

The Government must also give further efficacy to UK copyright law by requiring AI firms to be transparent about the methods by which copyright works are scraped, and the works that are used in AI training (and other uses such as fine-tuning and retrieval-augmented generation). This will incentivise AI firms to license data in accordance with copyright law and allow rightsholders to seek redress only where necessary, without having to resort to costly litigation merely in an attempt to discover if their works have been trained on or not.

The Government should also make a clear statement that AI firms wishing to deploy their models in the UK must train their models in accordance with UK copyright law, even if the training takes place elsewhere.

Ultimately, the Government must aim for a sovereign UK AI industry which partners with rightsholders to drive growth for businesses, consumers, and citizens. This will not be achieved by providing unchecked market access or free data for predominantly US-owned tech firms, but by creating the conditions for the UK to play to its strengths. To this end, we urge the Government to engage with UK AI firms – such as those in membership of this Coalition - and UK companies seeking to use AI, to ensure they understand the interests of the entire AI ecosystem, not just a handful of the biggest firms.

2.              What practical and technical mechanisms for (a) rights reservation and (b) transparency would provide rightsholders with sufficient control over their work while being proportionate and administratively reasonable?

a)              What, if any, legislative changes are needed to support rights reservation and transparency arrangements to function effectively?

The EU opt-out model – a rights reservation model by another name – has failed to support a strong licensing market. Indeed, the EU DSM Copyright Directive, with introduce a text-and-data-mining exception with an opt-out, was passed before the explosion of generative AI. Alex Voss MEP, who helped develop the Copyright Directive, has said that the exception was intended to have limited private use, and that its use for mass AI scraping is a “misunderstanding”.[1] Clearly, an opt-out model is not the best route to incentivising AI firms to license copyright works, or giving rightsholders sufficient control.

It should also be noted that a change in copyright law would generate years of further uncertainty for rightsholders and AI firms, with any exception certain to be tested for years in the courts.

Transparency mechanisms (which, for clarity, will not impact underlying copyright law, but simply give efficacy to it) are unlikely to be effective without primary legislation, as there are strong incentives for AI firms not to participate in a voluntary model. It is also clear that the Government should not simply replicate the approach of the EU, which fails to offer any granularity that would be useful to rightsholders in determining whether and how their copyright works have been used.

Transparency is necessary in two key areas in relation to AI training: (a) how copyright works are scraped by AI firms; and (b) which specific works are used in AI training. As set out above, this will incentivise AI firms to license data in accordance with copyright law, and also allow rightsholders to seek redress in cases of infringement. The reduction in informational asymmetries will also help rightsholders to negotiate for the true value of their works.

We note that transparency is relevant to a broad range of AI policy issues – privacy and data protection, AI safety testing, preventing illegal and harmful content from being used to train models, and allowing bias and prejudice to be identified. The Government should rule out a commercial text-and-data-mining exception at the earliest opportunity in order to spur licensing deals, and can then consider the issue of transparency holistically (for example, as part of consultation around any future AI Bill).

4.              What are the opportunities for the development of a UK licensing market that would benefit rightsholders and AI developers, and how can these be maximised?

a)              What role could the Government’s proposed ‘creative content exchange’ play in this context? 

The UK has an immense opportunity to exploit its wealth of high-quality data in the age of AI. AI firms will ultimately negotiate for the content that is the essential fuel in their models if the Government provides market conditions that incentivise commercial negotiations. As set out above, swift confirmation that the Government will not move forward with a text-and-data-mining exception is the crucial first step in spurring a dynamic licensing market, and a change in copyright law would only deepen the uncertainty and dampen the market for years to come.

We support the broad aim of increasing licensing by AI developers. This aim is currently being achieved by rightsholders who are licensing through existing commercial frameworks and innovative new voluntary models. Policy proposals such as the Creative Content Exchange could potentially collaborate with existing voluntary, industry-led licensing schemes to further increase and promote licensing. However, we are clear that the Exchange will not be an appropriate marketplace for all creative subsectors or licensing scenarios, and the Government must ensure that rightsholders remain free to license their works in whatever manner they see fit. More broadly, the Government must recognise that licensing initiatives like the Exchange will inevitably lack efficacy if they allow the uncertainty over policy direction to persist. It is transparency, allowing rightsholders to understand the use of their work and negotiate a license, that will be crucial in giving efficacy to existing copyright law – the Government’s first priority must be creating the right market conditions, rather than creating a marketplace. To focus on the latter without fixing the former would be to put the cart before the horse.

5.              What lessons can be drawn from the approaches taken to balancing the interests of rightsholders and AI developers in other jurisdictions?  

The key lesson to draw from approaches taken in other jurisdictions is that weakening copyright law is unnecessary to support AI development and will not support a dynamic licensing market. Combined with a series of significant court cases which are being won by rightsholders, evidence from other jurisdictions makes it clear that the UK would not be an outlier in maintaining its existing copyright framework.

The Australian Government has recently confirmed that it will not weaken copyright protections, noting that its decision “is providing certainty to Australian creators”.[2] The UK Government should mirror this approach. Following this decision, the Australian Government is convening a Copyright and AI Reference Group to explore avenues for less costly enforcement, and ways to encourage fair, legal avenues for using copyright works in AI – this demonstrates that a far more productive conversation on avenues to support the creative industries and AI firms can be opened up if a damaging change in copyright law is taken off the table.

Having passed the AI Promotion Act to make the nation more AI-friendly, Japan is now calling on OpenAI to refrain from infringing on Japanese IP. Cabinet Office Minister Minoru Kiuchi (responsible for IP and AI strategy) announced that the government had been in touch with OpenAI to ask them to refrain from infringing on Japanese IP, calling manga and anime “irreplaceable treasures”.[3]

As set out above, the opt-out in the EU Copyright Directive was developed before the explosion of generative AI. The EU AI Act’s obligation for AI firms to produce a “sufficiently detailed summary” - Article 53(1)(d) - of training data risks leaving rightsholders without the necessary information to guard against, or seek redress for, mass infringement. The EU AI Office’s ‘Template for the Public Summary of Training Content for general-purpose AI models’ – which sets out how AI firms should comply with the Act’s requirement – requires “disclosure of the main datasets that were used to train the model, such as large private or public databases, and a comprehensive narrative description of the data scraped online by or on behalf of the provider”.[4] Yet without information on the individual works used by AI firms, it is not possible for rightsholders to easily or effectively assert these rights, in negotiations or in the courts.

In the US – which is often cited as an example of a jurisdiction with far more permissive copyright law than the UK, demonstrating a need for the UK to change its law in order to be competitive – rightsholders are in fact winning significant court cases. Recently, authors were given a $1.5 billion settlement in a case against Anthropic, with rightsholders to receive at least $3,000 per title.[5] AI developers have claimed that AI training without a licence is ‘fair use’ under US copyright law, but it is increasingly clear that this is not the case, and is not a credible argument to weaken the law in other jurisdictions.

In the EU, whilst grave concerns remain that the opt-out regime has weakened rights whilst failing to spur a licensing market, it also appears that the opt-out does not allow mass scraping of copyright works for any purpose beyond initial training. In a recent landmark ruling, a German court found that OpenAI had infringed copyright, with works memorised in its large language model and reproduced in the model’s outputs.[6] The court accepted that training data can be embedded in model weights and continues to be retrievable from the model, and found that this does not meet the text-and-data-mining exception.

2 December 2025

5