Background
The new technology landscape emerging from the super rapid progress in developing AI, Generative AI (GenAI) and towards Artificial General Intelligence (AGI) has been dominated by hyperscale super-funded, largely US-based players. Despite the potential opportunities created by “distillation” from left field entrants such as DeepSeek, the major players are likely still to dominate.
Some commentators have thrown up their hands in despair and see this as the end of any chance to achieve a more open, widely distributed new technology ecosystem. They say the power of big compute and Large Language Models (LLMs) is such that no other companies could compete. They fear we are headed for an oligarchal technocracy which will display scant regard for legacy intellectual property (IP) regulation or any other interest that does not benefit the giant technology companies themselves.
Their fears are misplaced. The new global capability offered by Google, OpenAI, Meta, and Anthropic is on the one hand extraordinary in its ability to capture and draw on knowledge from a massive range of sources and is, on the other hand, still likely to be generic. The landscape is still open to new entrants as the development cycle continues to spiral upwards and specialist application areas open themselves for treatment. LLMs are less likely to be trained to become expert in more niche specialist areas of individual industrial sectors, such as aerospace regulatory requirements or IP rights in the creative industries. This is not because they could not be trained in those areas, but because the costs of training and sector systems integration are unlikely to yield sufficient levels of commercial revenue to the large players. As a consequence of this and the emergence of open source models, venture capitalists are beginning to identify specialist application areas as potential targets for investment. Any changes to IP policy should be cognisant of that shifting focus and not get in its way. Changes should neither hamper the ability of existing growth sectors to contribute to the economy, nor must they be allowed to inhibit cultural development by disincentivising creators and acts of creativity.
IP Rights Reform
IP rights-owning businesses in the creative industries are rightly concerned to regain the ground lost to GenAI companies who have exploited the world’s music, literature, games, photos and films to train GenAI platforms without regard for the rights of creators or rights owners. That wrong is obvious and needs fixing.
Historically we have seen remedies, either in court by creating precedents, or by governments changing the law. Commercial market-level solutions are also often achieved through negotiated settlements, agreed out of court where the terms remain opaque and not precedent-setting. Some constituencies in this debate might prefer that outcome. However, if we are to achieve a favourable shift in IP policy and legislation over time to reflect the new uses of content by AI, then we must make sure that those kinds of private deals do not prejudice the fair and open development of the market. It is therefore very welcome that the UK Government sees the need to work diligently, and at pace, to try to achieve a new kind of legislative IP environment designed for the longer term; an IP environment that will continue to incentivise authors, recording artists, photographers and fine artists in particular, as well as stimulate the emergent AI economy.
The fact is that as AI platforms continue to be developed, their level of opacity needs to be radically reduced. The sources of data that are included as constituent learning components in the output of content from GenAI platforms should be required to be made visible. Scientists continue to argue that the goal of AGI is a degree of autonomy or agency that goes beyond what individual prompt designers could achieve. This does not obviate the fact that an autonomous agent using its intelligence to make a new piece of content will do so under a lesser or greater degree of direction from a human, and base it on content upon which it was trained. The GenAI platforms must be required to invest in their systems so that they can render those sources visible and explainable. This will make a significant contribution to the global battle against misinformation too.
Reasonable parties agree that licensing content for AI training purposes should become a friction-free commercial service. Use of that content into newly generated material must be required to be traceable and paid for. If such a solution were designed in from the get-go, then Gen AI could become a valuable revenue stream for content creators and rights owners alike.
Regulators and legislators are right to reflect on the nuanced different positions of rights owning companies and individual creators. It might well serve both constituencies to have the GenAI platforms render their sources transparently and link them to their economic models. If they are not forced to do so by law then a potentially unholy alliance between the platforms and the rights holders might motivate them to agree out of court on, for example, a standard royalty pay-out based not on works used in music, literature or photography, but on more generic indicators such as market shares of the rights owning company. This would be cruder and cheaper to implement. It could form the basis of a deal between, for example, a major record label or a photography rights owner and a GenAI platform. The result of which would see the rights owner receive a major new revenue stream and the GenAI platform could claim that it was complying with a form of the law. The individual originator of the content, the artist, might receive nothing at all, since the record label could justifiably claim that it did not know the precise source origin for which it was receiving payment. This kind of “black box revenue” is a common feature in income streams in some existing collecting societies and needs to be avoided in any new framework for licensing rights for AI training. Such a scenario might suit the more powerful commercial players and further prejudice the interests of individual artists who historically have struggled to litigate for their own benefit in such a context.
A New AI IP right
In light of this, I do not believe that offering an exception to encourage AI training with an opt out right is the correct approach. Instead, I propose that a new statutory AI training ingestion right is created for the output of artists, photographers, composers and authors. This new right would be separate from legacy publishing and recording rights. This should be a new intellectual property right to allow or forbid the use of original content to be used to train algorithms and create artificial intelligence in GenAI and LLM platform environments. This should not be in the form of an opt out right, but rather a new statutory right that recognises that the function of ingestion is neither copying nor reproduction in the traditional sense and may therefore not be covered by the Copyright Designs and Patents Act 1988. Instead, ingestion leads to the creation of derivative works whose provenance may be traced. The obligation must be placed statutorily and actively on the technology company to reveal its sources in a transparent and auditable fashion – and to seek permission to make use of this right.
A new global rights body
Given the scale of the works to which these rights would apply, they would need to be enforced and administered by a new collecting organisation, which should be funded on a not-for-profit basis by the tech platforms. This new Collective Management Organisation (CMO) would need to operate from day one on a global basis and be granted global regulatory authority. The UK Government will need to use its international convening power to build a consensus around this approach. the. New CMO should administer permissions requests and royalties generated by GenAI. Those royalties should be paid directly to content creators themselves, not to corporate entities who have licensed traditional IP rights on an exclusive basis, nor to existing CMOs. Contractual terms that seek to expand exclusivity to future rights should not be within the purview of these new rights. Exclusivity should not extend to the new AI rights unless the individual creator has actively stated a willingness to extend. In addition, the new collecting and enforcement organisation must also establish a dispute resolution process which will be required to manage claims and counterclaims by writers and performers against technology companies and against each other. Having said that, once an acceptable form of royalty payment for content usage is fixed, the future (and hopefully past) use of licensed content will become a major new revenue stream for all classes of rights owners. Equally, the AI companies would be expected to incorporate these processes and costs into their own business models and pass on the costs of licensing to users, which would of course include creatives who use their services.
Stimulus for the UK AI economy
Quite aside from these distracting IP issues, there is a rapidly developing opportunity to create a new set of platforms and applications specific to the specialist needs of the creative, production, distribution and consumption value chains that are at the heart of the creative industries. The UK Government should not wait for the IP problems of attribution and recompense to be solved for GenAI before encouraging the development of this new field in which the UK could genuinely lead the world with significant yields in increased productivity and automation.
This new field will incorporate products and services which will also bring together the interests of the burgeoning creator economy, as well as the creative industries. The world of brand marketing has already extended its operations deep into the creator economy on the key three social media platforms YouTube, Instagram and TikTok. The creation of dedicated AI tools and services for content production, the ownership of bespoke AI services to create narratives, characters, and entire storyworlds appropriate for this kind of environment, as well as for more traditional linear content, will be key new ingredients of this sector.
With the UK’s global leadership in so many of creative industries’ sub-sectors, there is an opportunity for the UK to lead development and growth of this new part of the ecosystem. We need a national AI strategy to achieve that. We need clear programmes to grow the skills, develop the technical expertise and incentivise the commercial investment to make this happen here. The IP regime in the UK must be sufficiently forward-looking to embrace AGI’s direction of travel and ensure that our IP framework facilitates the right kinds of commercial incentives to motivate the new range of business opportunities arising from both the application and development of AI. This is also an important opportunity to enforce a better relationship between creators, artists, and publishers, producers and distributors, who for so long have accepted as unavoidable an imbalance of power that is reflected in unregulated trading practices and commercial contracts.
There is a new AI driven creative sector emerging where the UK can take a lead. It will embrace streaming platforms and content generation applications, avatar integrations and digital asset manipulation in production workflows. The companies that develop and run these will achieve increases in productivity and reductions in cost across content value chains, from scriptwriting to animation, mixing to mastering, set design to previsualisation, LiDAR scanning to location mapping. It is these new roles and new businesses that a reformed UK IP strategy must facilitate and enable. Many other countries are already building strong reputations in the AI space, France in particular around Mistral.ai, but the UK’s global leadership in so many areas of deep AI and the creative industries gives us a unique advantage.
Some of this progress is already visible. We have pioneering global leaders like Synthesia and Wayve, but we also boast smaller leading-edge companies like Figaro.ai which has been focussing on specific issues that face producers searching for music for use in film or TV. AI driven solutions are making music content more accessible and better differentiated. But investment for these kinds of businesses has been woefully slow. There is a paucity of public funding to help incentivise private money and an outdated IP framework that needs fixing. Uncertainty is hampering investment, and the proposed creation of exceptions with an opt out that could destroy projected licensing revenues would be a damaging misstep.
The UK is uniquely positioned to lead in developing AI for specific industrial sectors, starting with the creative industries. We must not delay building the new long term value components of this part of our industrial strategy, so the Government’s speed in issuing this consultation needs to be matched by an equally rapid but considered response. To complement this, we also need a funding strategy that is smart about how it uses small amounts of strategic public investment to attract substantial private capital to grow both the AI and creative industries in the UK.
© Jeremy Silver, 2025
All rights reserved
Dr Jeremy Silver
Dr Jeremy Silver is an Honorary Professor in the Computer Science Department at UCL. He is a Trustee of the British Library. He is Chairman of Figaro.AI, an audio intelligence company. He is a non-executive director of the Academy for Contemporary Music, a higher education institution. He is on the board of Studio Wayne McGregor, a contemporary dance studio.
He was previously CEO of Digital Catapult, Executive Chairman of Semetric and strategic advisor to Shazam. He was CEO of Sibelius Software and Vice-President of New Media for the EMI Music Group in Los Angeles.
His PhD in English Literature is from the University of London. His most recent book, Towards a Digital Renaissance, explores the complex relationship between investors, founders, technology and creativity.
Disclaimer: The views expressed in this blog are solely those of the author.
