Isabelle Doran, Anna Ganley, Tom Kiehl and Owen Meredith.

Q1 The Chair: Good afternoon. Welcome to this meeting of the Communications and Digital Committee. This is our first session of a new inquiry focused on AI and copyright, which the committee is launching this week. Our aim is to build on the previous work that the committee has done in this area and to complement the Government’s ongoing work on the topic following the consultation at the end of last year. We are very pleased to have four witnesses with us today who represent a range of different rights holders. Today’s session will enable us to discuss your perspectives on the challenges as well as questions relating to transparency, licensing, and management of AI outputs.

The session is being broadcast live and a transcript will be taken. It is the first session of our inquiry, so I would ask members to think about whether they need to declare any interests before they first speak in this session. I have no interests to declare. I am Baroness Barbara Keeley and I am the Chair. I will start by asking you to introduce yourselves and the organisations you represent, starting with Tom Kiehl.

Tom Kiehl: I am the chief executive of UK Music. UK Music is an umbrella body for the commercial music industry. We represent 10 organisations that represent both business interests and creator interests across the music industry. We are there to campaign, collaborate and champion on behalf of them. The members include BPI, Music Publishers Association, PRS for Music, PPL, AIM—the Association of Independent Music—Ivors Academy, Musicians’ Union, Music Producers Guild, Featured Artists Coalition, and the Music Managers Forum.

Isabelle Doran: I am the CEO of the Association of Photographers. I am also vice chair of the Creators’ Rights Alliance as well as a board director of the British Copyright Council. The AOP represents about 3,000 photographer members, mostly freelance individuals. We also represent a number of agents who represent photographers, as well as university courses and students.

Anna Ganley: I am the chief executive of the Society of Authors, which is a special registered trade union for writers, illustrators and literary translators. We represent 12,500 members. I am also currently chair of the Creators’ Rights Alliance, which represents half a million individual creators, alongside Isabelle.

Owen Meredith: I am the chief executive of the News Media Association. We work with the vast majority of UK news publishers, ranging from national, regional and local publishers, in total representing about 900 titles across the UK.

Q2 The Chair: Thank you. I would like to start by asking our panel to talk us through the main challenges that generative AI poses for creators and the rights holders that you represent and what impact these have had to date.

Tom Kiehl: From the music industry’s perspective, our position is quite clear. The music industry is worth £7.6 billion to the economy. It generates £4.6 billion in exports. It employs over 216,000 people. Beyond the commercial impact that the music industry has, it has other profound impacts on society: the cultural impact, and the soft power impact in a number of areas.

The copyright framework is fundamental to the music industry, from both a business perspective in terms of the investment that goes into the sector but also the creators and the value that they have from their creations and their IP. The copyright framework is fundamental to that success, and that powers that £7.6 billion. I very much welcome that the committee has dedicated some time to this very important inquiry. We have seen earlier this year the profound interest and knowledge across your Lordships’ House on these matters, so I very much hope that the work you are doing will help to contribute to these further debates and discussions.

There are a number of particular challenges that we face in the music industry regarding artificial intelligence and the way it impacts. That said, before going into some of the challenges, there are some benefits to artificial intelligence, particularly on the assistive, creative side of the music industry or maybe the business efficiencies. The big challenges are faced in some of the areas around generative AI. In particular, we are thinking about things like the mass scraping of works without consent. From our perspective, it is very concerning that that has been going on and that AI systems and large language models have been trained on music. We have seen examples such as the Anthropic legal case, where particular examples have been given regarding prompts that have been put in and then generated—verbatim copies of Gloria Gaynor or Louis Armstrong lyrics. Louis Armstrong did not even write the song, but it has shown the way that the systems are being trained on the back of creators and rights holders.

There is another issue, which is competition with human-generated works. We have seen new pieces of work being generated and created in this way, particularly on a legal basis behind them, that are conflicting with legitimate sources. From a music industry perspective, we are striving to work with rights holders and creators to make sure that they get value for their works. For example, a French collecting society organisation and international collecting society organisations have indicated that there is a 24% potential loss of revenues by 2028 if things carry on as they are.

Generally, around consumer confidence, there has been a lot of talk recently about what we consider to be AI slop—inferior and low-quality AI-generated systems, particularly streaming models, being very prevalent. I think that Deezer, which is a French digital platform available internationally, has seen around a 28% increase in AI-generated music content uploaded on its platform. That is up from 10% earlier this year. There are some big, profound challenges that we are facing within the music industry.

The copyright framework is fundamental. We will presumably go into this in a bit more detail, but the issue around transparency is key for us. Unlocking some of those challenges around consent and getting more transparency is key to maybe some of the solutions that we need. We do not have that transparency at the moment, and that is a fundamental challenge.

Also, there are issues around particularly the way music works in the online environment. A song is uploaded to a digital platform but then it can go to another user-generated content service such as YouTube. It can manifest in multiple ways, so it makes it very hard for rights holders and creators to enforce their rights in that way. They are what we call downstream copies. In terms of enforcing our rights and the way they are scraped, it is very challenging. Added to this is that a lot of the training of the AI systems takes place outside, in maybe more liberal regimes. When you think about some of the proposals the EU has around an EU AI Act, ensuring compliance with UK law for anything is key to the fundamental issues.

At the end of the day, we must remember that in the music industry about 70% of the people who work in it are freelancers and self-employed. There are also a lot of small and medium enterprises. Their ability to navigate that process is very challenging. There are big players in the sector, which is important, but there are lots of smaller players, too. It is a very welcome inquiry and I congratulate you on your initiative in taking this forward.

Isabelle Doran: I would like to echo Tom’s points. Thank you very much for convening this opportunity to give oral evidence at this committee meeting.

To give you some photographer stats from our perspective, by and large members of the Association of Photographers work as freelancers, which obviously puts them in a pretty precarious place to start with, whether they are emerging talent or established professional image makers. They create both still and moving images, not exclusively just photography any more, and invest, of course, a significant amount of time, effort and money to produce their work and differentiate themselves in the marketplace. I want to underline that differentiation because it is crucial to their livelihoods in that respect.

While we represent 3,000 members, there are approximately 50,000 professional photographers in England and Wales. That is an unofficial UK ONS stat, but we have been able to at least try to measure the scale of the number of professional photographers. We know that there are 2.4 million creative workers in the UK. The average turnover for our surveyed members is around £105,000, which sounds like a lot but obviously that is before they take home any profit. That was up by 21% in two years when we did the survey, back in February.

However, what was very interesting was the statistic that also came out, where over 58% of our surveyed photographers had lost work to generative AI. At the moment, that is probably one of the highest stats that we have seen. It doubled from five months prior to that when we did a survey with our members, and again we can see that correlation with the arrival of generative AI.

An average photographic shoot will cost anywhere between £29,000 to £100,000, so a lot of money is involved which a photographer will invest in their work. The average loss value transfer in the survey that we did was approximately £43 million, which is the equivalent of about £14,400 per photographer. On each photographer scale it does not sound very much, but I have some anecdotes. For example, “I have seen a drop of £100,000 but that is for full production loss. The actual photographic part of this was about £20,000. The rest was what we supplied with the package: studio equipment rental, post-production, a producer and a stylist”. Again, when we are thinking about a photographic shoot, it is not just an individual, it is everybody who works with that person.

In this brave new world of AI, we are specifically concerned with generative AI, which for us mimics human originality, taking away the heart and soul of what it means to be a photographer or image maker. I have a few other anecdotes from our members. “One of my most iconic images has been compromised. Generative AI undermines my ability to control the use of my work and therefore my IP income is reduced to 0%.” We also know of established photographer agents who have closed their businesses owing to generative AI’s arrival in the marketplace. We know of several higher education universities that have closed their professional photography courses, instead amalgamating courses into multimedia ones and losing out on the professional support they would have been getting from us. Of course, this is despite the fact that the Migration Advisory Committee recently published its temporary shortage list, which includes photographers. It is a bit of an unusual outcome for that.

Specifically on the challenges, as Tom was saying, consent must come first. This is the challenge we have. Our UK copyright framework is the gold standard. I do not think that there is any shadow of a doubt about that. Copyright is a human right to protect livelihoods. Creators have little knowledge or control over the exploitation of their work—whether it is published online—particularly by big tech, which, it seems, flouts UK copyright law by training overseas. Machines cannot yet unlearn, and when it comes to images, we are talking about diffusion models as opposed to large language models. There is no control for rights holders over the downstream use of their work. Many of our photographers are on commissioned work, so they will be working with news publishers, music companies, all sorts of different clients. Essentially, they have very little control once they have handed it over.

With photography, what we are talking about is the sheer scale of the amount of works. To again give you some context, the average number of images on a photographer’s website and what they license is approximately 11,000. However, the average size of their archive is about 40 times that. When we did our survey, it was around 440,000 images. When we are talking about tools and ways of being able to mitigate particular risks, the burden always falls on the creator to have to implement a lot of these things, which can be very costly. There was a period of time when we had to add in metadata to that number of images. It is a significant cost and time resource.

This is against the backdrop of the amount of AI-generated images, which is estimated at 15 billion, with an average output of about 34 million synthetic images per day. Where we are talking about challenges, it really is trying to find a commercial infringement among those outputs. It is like finding a needle in a haystack. I do not think that there is a more appropriate analogy than that. Even if creators find infringements, the perception is of the high risk of taking a case to court; the cost and time involved in litigating against, for example, big tech, potentially losing and then setting off, for example, wide-scale “in the style of” infringement—and I want to underline the “in the style of” point, because that has become a significant burden for creators. Part of the displacement that is happening with their jobs is the fact that they have no opportunity to protect the style of their work. One of the things we would be asking is whether it is time for a UK personality right that sits under trademark, rather than copyright.

Anna Ganley: Building on what Tom and Isabelle have said, I will go through all the stats and figures that relate to authors. However, I want to take a slightly different tack, bring it back to basics and tell you all a story, since I am from the Society of Authors. It is a story that you will all know well, of Robin Hood and his merry men, only in this story it is the creative industries that have had their intellectual property stolen and used by rich tech companies. This industrial-scale theft of their copyright-protected works feels like a modern-day tale of Robin Hood, as it has led to an unprecedented campaign by the UK’s creative industries, which are all united on this issue, against the theft of their intellectual property.

We know that all the creative industries have banded together to fight this unlicensed use of their work to ensure that there is future regulation to hold rich tech companies to account. The challenge, as we know, is that big tech has done this. They have used copyright-protected work without permission, without payment, and they are getting away with it. That is making our members, our authors, very angry: they feel helpless and hopeless because big tech is seemingly above the law. There is little to no transparency, as my colleagues have said, where big tech have used authors’ and creators’ works, and transparency has to be a key part of any solution going forward. The challenge is that our members are looking for compensation for works that have already been scraped for these large language models. Compensation retrospectively is one part of it, and then going forward is the regulation to future-proof an AI industry that is embedded on trust. You need transparency to have that, and we do not currently have it.

We are seeing with the Bartz v Anthropic case that transparency is possible. Tech developers can disclose their data. Without this visibility, creators cannot know where or how their work has been used and therefore they have no redress for any unlicensed use. We are building on the Bartz v Anthropic case to try to get transparency across the board.

In our conversations with tech developers, they say we need to keep a human in the loop when it comes to using AI systems, but we also need to protect the human in the loop when it comes to the development of AI systems. As Isabelle said, copyright is a human right and we need to protect that. Copyright, in case you need a quick recap, is how creators earn from their creative endeavour. It exists to allow creators to control how their work is used and this is how they monetise their work. In the publishing industry we already have very good licensing frameworks, which allow buyers to license the use of those copyright-protected works for a range of uses.

Going back to the tale, this is a tale about jobs and money in the creative industries, and we need to protect and safeguard them. Authors are not against AI but they are against their work being stolen. It has been proved through The Atlantic article that millions of copies of books and journal articles have been scraped by pirate libraries, and then trillion dollar tech companies have chosen that quick and easy way to train their large language models rather than licensing in the lawful way.

As Isabelle and Tom said, we have a gold standard copyright system. We do not need any changes to the copyright system. We need regulation. We do not want copyright to be diluted. What we want is for big tech to do the right thing and to copy right. We have seen recently that the world’s most highly valued companies are tech companies—no surprise: NVIDIA, Apple, Microsoft, Alphabet, Amazon. They are getting richer, and individual creators—we represent over half a million of those creators—are getting poorer. To give you a stat, in the latest Authors’ Licensing and Collecting Society survey, the median income for an author from their book sales was £7,000. Expecting individual creators to chase the likes of Meta and big tech over the unlicensed use of their work is unrealistic and puts an impossible burden on the individual creator. We need enforcement and we need regulation.

We are very glad that this inquiry is happening, because it felt like something was wrong when the Government published their AI Opportunities Action Plan before the consultation had closed earlier this year. This seemed to us to indicate that there were some predetermined outcomes. Creators would like to be part of the process. We are here and ready to talk to technology companies to make sure that there are solutions that work for creators. Creators are the cornerstone of the creative industries, so without creators these large language models have nothing to train their work on. So let us do it the right way; let us do it the legal way.

With the Government refusing to implement the AI data transparency measures, it just feels like there is something wrong when the whole of the creative industries is aligned and saying we need regulation and we need transparency. It is unprecedented, so we want to make it fair. We need to put this right and we would love your help to do that with us.

I can run through some stats before I hand over to Owen. For authors, their incomes have been in decline for the last 16 years, with a 60% decrease in their earnings. We found when we surveyed our members in 2024 and 2025 that 86% of authors said that AI advances had already decreased their ability to earn. Our literary translators are the canaries in the coalmine. They are seeing their livelihoods decimated by the fact that they are now being asked to correct machine-translated text rather than translating from scratch, which is often quicker, but they are being offered less. The creators that we are seeing most affected are definitely copywriters, translators and script writers. As a representative from the Creators’ Rights Alliance, I also represent audio narrators, voiceover artists and actors, who are also seeing their incomes, voices, faces and likenesses being taken unauthorised.

I have a few more stats and then I will hand over. Over 57% of authors no longer consider that creative work will be a sustainable career. Over a third of our illustrators have already lost commissions to AI-generated content, with losses across the industry of around £9,262 per creator. Across the board, 93% of our members have said that AI is an existential threat to their ongoing employment. We simply cannot champion global innovation by exploiting the very creativity upon which it depends.

I will finish by telling you a little story. As we look ahead to the national year of reading, what I would like to know is how we will inspire the next generation of readers and writers. As a sector, we have come so far with diversity and inclusion, but with every day that passes individual creators are losing money, reconsidering their options and exiting the creative industries. I grew up in the north-east of England in a little village called Whitburn. Both my parents were teachers and we had no family or friends in publishing. We had no family money to support me into a creative career, but a love of reading and books and our local library got me into working for the Society of Authors, where I help authors every day. Without government support on this issue, we risk a huge backwards step and narrowing of the creative industries back to the preserve of the privileged few who can afford to create without having to pay the bills each month.

Creative careers are fast becoming unsustainable, especially if government continue to allow the unauthorised use of copyright-protected works by global trillion-dollar tech companies to train their commercial products without paying for their raw materials. If we lose diversity within our creative industries, important stories will go untold, leading to a narrowing of voices and the further diminishment of empathy, critical thinking and ultimately our shared humanity—and we must hold on to that.

Owen Meredith: I am conscious that the committee has a lot to get through.

The Chair: We have questions about transparency and options for licensing and other things coming, although they have already been mentioned.

Owen Meredith: In the interests of the committee’s time I will not repeat what colleagues have said, but I wholeheartedly endorse the vast majority of it. Certainly, from a news perspective we have seen content scraped without transparency, without consent and without reward, and that is fundamentally unfair and, in our view, illegal.

A lot of the debate that we have had over the past 18 months or two years since generative AI exploded on to the scene has been almost pitching AI versus rights holders. I think that we need to fundamentally move away from that. The UK should have confidence in itself in the strength of both those sectors. The Government in their own industrial strategy highlighted both sectors, but so far seem to be leaning very much towards supporting the technology sector, to the detriment of the creative industries and rights holders. I think that we need to be more confident in our own ability and not be shy. The UK can have its own legal system and exercise its own jurisdiction in a way that supports both sectors to grow alongside each other, as they have done over the last 10 years with the creative industries being the fastest growing sector in the UK. There are plenty of examples where we have demonstrated a UK-first agenda and approach, and I think that we can continue to do that in this space. I am sure we will get on to some more solutions as we explore that through the committee.

From a news perspective, often, news content is thought about in the context of text and photography and imaging that sits alongside that text, but increasingly important for the sector and its sustainability in the long term is audio and video content, and new forms of demonstrating that news content, all with the core underpinning of being editorially controlled, legally accountable and fact-checked content. That is what I would hope the UK Government and regulators want to support, given all its democratic and societal benefits and values.

I think we can get to a point where a licensing system flourishes, in order for monetary value to flow back to newsrooms. This is particularly the case when we are talking about RAG rather than the training aspect through large language models, although both clearly have a value for news content. RAG is where models retrieve live information essentially from news publishers’ websites to complement what the model has pulled from its training history, to validate up-to-date current facts. Without access to current news content, those models will become increasingly unsustainable and unreliable for consumers, to the detriment of the consumer experience. It is in developers’ interests to ensure that there is financial sustainability for the news sector.

I will close on this point. The other aspect that we have seen is clearly enforcement. I would agree with colleagues that we do not need to fundamentally change the UK gold standard copyright regime. I think that it is fit for purpose. However, speed of enforcement, where you can demonstrate that copyright has been breached, is the thing that is fundamentally slowing down the ability of creators to act. The Government in some ways have been complicit in creating uncertainty in the market through the way they have approached their policy agenda in this area. It has created further uncertainty and has allowed developer firms to exploit that uncertainty to their own benefit.

The Chair: Thank you. That is a fairly extensive opening. We will come on to questions about transparency now.

Q3 Baroness Wheatcroft: Thank you for explaining the depth of the problems and the hits it is having on your sectors. What I would be particularly interested to know, though, is how you could have a system that works for individual photographers, authors and so on. What information would they actually make use of? We will start with Isabelle.

Isabelle Doran: To give you some context, when we went to our members to ask them about transparency, over 98% wanted AI companies to be transparent about using their works. When we are talking about what we want to see, essentially, we are talking about traceability. Particularly across the data-processing pathways through which models are using creative works to shape the downstream outputs, we want them to disclose what was used, how and where those works were obtained and particularly the model weight. Over the years since we have seen and understood more about generative AI, we understand that model weights are particularly pertinent to how a programme will operate.

In some respects what we want to see is much like the current food standards process. We are not necessarily talking about revealing trade secrets. What we are talking about is the food standards process, which we are all pretty familiar with. One of the key things about it is the provenance of those works and how they move through the chain, so that a consumer at the end of it knows and understands what they are getting. They can see the ingredients but they do not necessarily know how it is made as such.

To give you some examples, when one open-source image-based dataset was made searchable, which was quite early on, a lot of our members did explore that. Some did not because the burden was too great given the number of images, but it was an opportunity to have a look and find that works did exist. The challenge they then had was what to do with that. The issue is that machines, as far as we are aware, especially with diffusion models, cannot unlearn. Despite the promise from some of the tech developers that, “Okay, we have accepted that you want to opt out your works”, which, again, is a burden, there is no proof of that. There is no proof that those works are actually removed. In many respects, there are papers that suggest that, even if you try to suggest to the machine to not output those particular works, it is a bit like trying to unlearn the alphabet.

Baroness Wheatcroft: The ultimate aim for your members, presumably, if they are being practical, is to get some remuneration.

Isabelle Doran: Very much so.

Baroness Wheatcroft: How are they proposing to do that?

Isabelle Doran: I think that this goes to the heart of the problem. While we are asking for transparency, one of the biggest challenges we have had, and we have seen it time and again, is the fact that big tech will not come to the table. They may deal with certain representatives who they feel perhaps have those rights, have the works, the data as such; but for individual creators what tends to be the best mechanism is collective rights management—collective management organisations that have mandates from quite a lot of creators. We are very aware that there are visual arts collective management organisations. There is one for authors, publishers, and so on. Their mandate is licensing, so they are set up to license on behalf of individual creators.

Probably the best example that we know of is education licensing. Education licensing was extremely complex when the Hargreaves review was first initiated. That was one of the elements that came out of that, and the copyright hub was explored. There were a number of reasons why it failed, but one of the key reasons was that big tech refused to adopt the provenance watermark that was proposed. It was proven to work, it was endorsed by picture libraries and agencies, but it was not carried through.

You will hear a very similar comment about an initiative called the C2PA. I will not go into what the acronym means, but again, it was born out of news provenance in terms of text and images. You will see that a lot of big names have adopted it, but it is not being utilised in the way it should be to label outputs. I know that is one of the questions or areas that you want to explore.

It gives you some idea of the fact that, while we are all willing and keen to get involved and to negotiate and discuss transparency and licensing, we need two parties at the table, essentially. The challenge we have is that big tech just do not want to be there.

Baroness Wheatcroft: That is very helpful. Anna, presumably you are finding exactly the same.

Anna Ganley: Building on what Isabelle has said, in the publishing industry we have seen examples of licensing, whether it is the HarperCollins deal or the academic and educational publishing licence deal. For anything that is not online, big tech have agreed deals with publishers to access datasets that they cannot scrape from the internet. So licensing is happening.

Baroness Wheatcroft: That is with the publishers rather than individual authors?

Anna Ganley: This is a right that rests with the authors, so it is subject to the author’s consent, to a separate agreement, to new terms being agreed. It is a one-off licence fee in perpetuity because once it has been learnt it has already been ingested by the model. That shows that licensing is happening and can happen, either directly with the publisher or through a collective management organisation through a blanket licence. There are licences in existence, but we need big tech to come to the table and have those conversations about how we can make it work per sector.

Q4 Baroness Wheatcroft: That is very helpful. Can I come to you, Tom? The music industry is slightly different, and you represent several organisations rather than individuals. Do you think that it is possible for individuals to protect themselves under this system?

Tom Kiehl: I do. Basically, we need to get to a certain point. The fundamental issue is that transparency is key to enforcing copyright law. Without that, you cannot have that licensing arrangement in place. I think that there is a big opportunity for the UK in having a gold standard around transparency. When you look at other provisions in other jurisdictions, the EU, for example, there are commitments but it is not sufficiently granular. Within the music industry, we have metadata and identifiers such as International Standard Musical Works Codes and International Standard Recording Codes. Each song can be tracked and a piece of musical work can be tracked in that way. Time periods and copying needs to be logged as well as part of that transparency process; what territory the copies were worked in, that international jurisdiction, and the legal basis of it. Those are three or four areas where the transparency model can work within the music industry.

On the broader point of the difference between the music industry and some of the other representatives around the table, again, we have experience of both direct licensing and collective licensing within the music industry. For 200 years we have been licensing music, whether it is sheet music, the gramophone, CDs, and in more recent times the digital innovation of streaming models, downloads, TikTok and YouTube. We are very good at it, both as an industry and as a country. We have some very sophisticated licensing organisations like PRS and PPL, which we represent in our membership, which collect money and distribute—PRS distributed £1 billion—to individual rights holders and creators.

The framework and infrastructure within the UK music industry is there to develop, and getting the transparency in place is key. There have been some developments over the summer; for example, independent distributors represented by a body called Merlin have done agreements with AI companies such as ElevenLabs. That is a positive step. Last week, there were reports of Udio, which is an AI service, coming to an agreement with Universal Music Group. I do not have details of the agreement that was made, but that shows that there is a developing licensing market. We just need more transparency.

That goes to my earlier point about transparency being key for individual creators and smaller businesses, as well as all businesses more generally. We depend on that in order to have that negotiation. I think that we are at the beginning of a process. That is why what happened was very disappointing. The Government earlier this year talked about the need to have a sector plan, which we are very supportive of as a sector, and to invest in the music industry, saying that we are a high sector for growth, which is very welcome indeed. But there were proposals at the beginning of the year to change copyright law and bring in exceptions to copyright. Exceptions to copyright occur when there is market failure, but there is no evidence of market failure at the moment. The only failure in is getting the AI companies to come to the table. That is the fundamental challenge that we are facing.

Baroness Wheatcroft: Owen, presumably you can get the big companies to come to the table?

Owen Meredith: Some of my members have the ability to negotiate directly with larger firms, but I represent pretty much the entirety of the UK news publishing market. There are a lot of independent local publishers, particularly small but even larger group publishers, which will not be able to negotiate. I think that we can come up with, and the industry is fully prepared to come up with, solutions that provide developers access to that content. I think that this committee has looked at issues around local news in the past, and this is a rich opportunity, potentially, for the local news market. If you want to know what is going on in Maidenhead, Newport, Aberdeen, wherever, there are one or two authoritative sources of information that is being gathered and authenticated on the ground, so that is of extreme, premium value to a developer of an AI model that does not have any other source to get that information. The publisher needs to be able to be rewarded for that.

When it comes to transparency, there are probably two aspects. The first is how content and data is being collected, and part of that requires a register of bots so that we can understand what crawlers are coming on to sites, identify who the owners are so that we are able to record that, and decide which bots and crawlers we want to block and for what purpose they are coming on. The second aspect is what data they are taking—so transparency over what data is being collected and again for what purpose, and a level of granularity that allows us to audit that process, particularly where licences do exist. It is a common practice where content is licensed that an audit mechanism sits behind that, so you can validate that the purpose you have licensed content for is being met.

Q5 Baroness Wheatcroft: Do you think that there should be a new regulator or a body that compiles the register and monitors the register? Otherwise, who should be doing it?

Owen Meredith: I do not take a strong view at this stage as to where that regulatory responsibility needs to sit—whether it needs a new regulator or whether there is an existing regulator that can fulfil that function. I think that is probably not one for us to decide at this point in time, but the idea that a regulatory backstop sits behind a bot register seems a very sensible one to me.

Finally on that point, you asked about how you get reward. There are increasingly market-driven technology solutions that allow publishers to control how their sites are crawled, to control on a granular level which bots are potentially coming on to their sites, to authorise them to do so for specific purposes and then to monetise that purpose—in other words, to produce a licence in real time, similar to the advertising market real-time bidding and the way programmatic advertising works. I think that technological solutions will come forward as soon as the regulatory environment creates the fundamental aspects for a market to be created. I am confident that entrepreneurial types will come forward and fill that gap.

Q6 Viscount Colville of Culross: I want to go back to the issue of transparency. Owen, you have just been talking about a register of bots. One of the things that interests me is how you can get to the original provenance of the data. We have been told that that data is transmogrified when it gets into LLMs. It is transmogrified again when it gets to the adaptation period for the end-use models. Is it possible, do you think, to get a complete list of what that data is? We have been told that, even if it is—for instance, one of the suggestions has been that there has been complete URL transparency on what data is used—that would be such a heavy burden on small AI developers that it would thwart their growth. Do you think that it is possible to go beyond the bots and get the original data labelled, and for that label to be maintained all the way through the process to the end user?

Owen Meredith: My understanding is yes, in short.

Q7 Viscount Colville of Culross: Isabelle, you talked about C2PA. How effective has that been in getting the original data provenance labelled and for that to be effective in trying to find out how your members’ content has been used? If it is effective, why has it not been rolled out further?

Isabelle Doran: In many respects, I would say it is probably ineffective. It is an initiative that a number have signed up to but in many respects it is machine-to-machine readable. To the previous question you were asking, when we are looking at how these uses of our works can be identified, it is machine to machine. It does not necessarily have to be a human being that requires it. The burden is much more on the creator in that respect, because they are literally having to gather their works and be able to process them through a particular system. If you are dealing with thousands of images, that is a real challenge.

Where C2PA is mentioned is really about the fact that it is an initiative that does not have any teeth as such. I suppose it is essentially giving your word, saying, “I will do this” but there is only, I think, one social media platform that utilises it and shows an identifier that says this work has either been created by a human or is synthetic. This is the thing: there is no legislation behind this that is mandatory and says that they need to be transparent, that they need to label this work, that they need to use these initiatives—not just saying “I am part of something”, but using it so that it is easier, as Owen was saying, to bring new technologies that can identify. One of the new technologies that sprung to mind was a company called Vermillio, which said it is able to identify works within an output, which is extraordinary. We know that technology eventually does catch up. The challenge we have is that in the meantime the pace at which the loss of work is happening is far faster and greater in that respect.

There are technology opportunities. The challenge is always back on the creator with the cost of implementing those technologies. We can go to Vermillio and say, “It is great you are able to identify those works, but what will it cost me to take out a subscription?” These are the practical challenges that often fall to creators to try to manage.

Viscount Colville of Culross: Would you like to see legislation imposed in order to make sure that—

Isabelle Doran: Certainly on transparency, and transparency goes much wider than just for the creative industries. Transparency is needed for data privacy. It is a societal request, in that respect, that we have. We should all be pushing for transparency on what goes into these machines, how our data is being used and how it is then further exploited. Without that, in many respects, we are still on this hamster’s wheel that we cannot get off or cannot move forward from.

The Chair: We need to move on now. I will come back to you later. I want to come to Charlotte now.

Q8 Baroness Owen of Alderley Edge: Given that it is the first session, I am declaring an interest as a guest of Google at its Future Forum.

Could we go over what the opportunities are for the UK to develop licensing markets for AI use of creative work that supports both rights holders and developers? Could we touch on what licensing arrangements you see as most workable for rights holders in the sector? Finally, could you also cover the Government’s proposal for a creative content exchange and whether it would facilitate licensing of creative works as training data? Do you see a good role for this?

Tom Kiehl: On the licensing, I mentioned that a bit earlier in one of my wider remarks on transparency. We do have a very strong licensing infrastructure in the UK. We are in the very early stages but we are licensing and I think that is important to recognise. The big concern was that we cannot bring in legislation that weakens copyright at this point. It is incredibly challenging for us to have that situation.

If I could go into more detail on the proposal around a creative content exchange, it is an interesting concept and very early days. It was part of the sector plan that was published just before the summer. The key is making sure that there is a voluntary approach. One of my messages around this would be to remember that about 10 years ago we were in a relatively similar discussion. The Government of the day were talking about a digital content exchange, which sometimes people refer to as a copyright hub. That got so far, but then it did not get further. I am not saying that this is necessarily the same thing, but when government goes into these areas and we want to address them constructively, maybe look at what happened 10 years ago, too, and make sure that we learn some of the lessons from that, instead of reinventing the wheel.

Isabelle Doran: When we asked our members about licensing, we had, as you would expect, quite an overwhelming percentage reply. Over 97% of photographers wanted to be compensated for past infringements. Again, it is important to underline that, essentially, before we can move on to future licensing, we need to redress what has happened in the past and not just sweep it under the carpet. We have programs that are competing with photographers and other creators, so compensatory licensing is essential to be able to move forward. In order to bring those creators along with us, they need to have the trust back in the system. Of course, this is about jobs and money at the end of the day, and creators are usually left out of the picture until right at the very end.

When we are talking about licensing opportunities, again, I have mentioned collective rights management. It is likely to be the best opportunity for individual creators because they do not have the opportunities a big organisation has. For example, not the Getty Images case, but Getty Images is a large organisation that has had the opportunity to go to court or to agree licences. It just last week agreed a multi-year licensing deal with Perplexity, for example. Individual creators or even representatives like us were not at those tables. We are not able to have those discussions. Collective rights management has that opportunity, but part of the challenge there—and we have heard behind closed doors about why this is not happening faster and right now—is the perception that CMOs do not have the data. One could argue that the reason you are seeing publishers with deals is that they have content; they have the images and text—the works—whereas collective rights managers only have a mandate. They still need to come to us to be able to have access to those works. In many respects there is almost a chicken and egg situation where they have said, “We have scraped everything; we have everything”—well, they assume they have, but I have mentioned the size of the archives—but they also do not feel that they can license through collective rights management, which has the mandate. Again, it comes down to, “Hold on a minute: should we not surely be talking and bringing those two groups together to agree those licences?”

Just to give you one other point on the creative content exchange, at the moment it seems it is mostly a handful of cultural heritage organisations that have been approached, so no creators have been approached. We know that at this stage it is embryonic, that there is a way to go. But I want to underline that a lot of cultural heritage organisations have a lot of in-copyright works, not just works that have fallen out of copyright or are public domain works not publicly available—there is a difference—so they are merely custodians of those works. In that respect, it means that they still need to get consent from rights holders and creators. Of course, I have already mentioned the copyright hub and the fact that big tech just simply did not want to get involved. I think that we are coming up against the same problem each time.

Anna Ganley: On the creative content exchange, we have not seen enough information yet. For the publishing sector, there is not an issue that it is there to resolve. It feels like duplication but we have not seen enough information. It is good that it is another option for structured licensing, but we already have various licensing options within the publishing sector.

Owen Meredith: Just to be brief, on the creative content exchange I agree with what others have said. I echo the point I made earlier. I do not necessarily think that the Government need to create the market; they need to create the conditions for the market to develop.

The other aspect of this is clearly extraterritoriality and how we apply British rules. I cannot think of any other supply chain where goods are stolen in the UK, exported to be parts of a finished product, and we then allow that finished product to be imported and sold in the UK market, exploiting the goods that were stolen from the UK market. I think that it is quite simple.

Q9 Lord Knight of Weymouth: Very briefly, following up on Baroness Owen’s questions, Isabelle, you talked about the CMOs. Is it fair to start to wonder whether there is a collective failure on the part of the collection societies—that they have not responded quickly enough and in an agile enough way to the threat we could see coming down the track from big tech, and which has arrived and caused huge problems to the people you represent?

Isabelle Doran: It is a good question. I would not necessarily agree that they have not acted fast enough because they have been in conversation with us from the start in that respect. We have been discussing it with them for quite some time. The challenge they have is not being met. The challenge they have is the fact that the doors are not opening.

Anna Ganley: We wrote to over 70 tech companies, the major tech companies, to ask them and we did not get an acknowledgment that they felt they needed a licence. They felt they had a lawful basis upon which to use the data that they have scraped. We did reach out through the Creators’ Rights Alliance and the Society of Authors. We wrote to them all and we got very few responses. Most of them said that they felt that they already had a basis upon which to use these materials.

Q10 Lord Knight of Weymouth: Thank you. Let me come on to the main theme, which is to think about the outputs from AI. You could in some ways argue that for the content that has already been created the ship has sailed on trying to monetise the point of training, and that therefore the desire, which is perfectly reasonable, for past infringements to be rewarded needs to look at the output and whether it is possible to monetise the output of AI in order to get some reasonable recompense. I am interested in whether that is possible and what legal and regulatory framework changes could be made in order to make that possible, so that rights holders are recognised and consumers can trust what they see and hear.

I suppose a starting point for me is that a couple of years ago, I was in Silicon Valley with the committee and we met one of the major investors in big tech. They said to us, “Well, look, the fact is that humans consume all sorts of content and they are inspired by that content and then they create content and the only point at which there is a problem is when they plagiarise that content and reproduce it accurately. Why change anything? What is the problem in output terms? If the AI reproduces someone’s content, then that person should be recompensed for it. If it is inspired by content, is that not just creation?” How would you respond to that?

Tom Kiehl: I do not think it is inspiration. This is copy; this is pure theft. This is the situation we are dealing with. Copyright protects the right of a copy, and when you have systems trained on material you are ripping off the copy.

Q11 Lord Knight of Weymouth: Thinking about a technology like the one Isabelle described—Vermillio, which identifies the works in an output—Ed Sheeran has had two court cases that I know of where he has been accused of plagiarism and, happily for him, he won. Is that not a danger for humans: that the same technology will be used on them and they will find themselves in trouble—that the things that have inspired them as creatives will then come back to bite them?

Tom Kiehl: We have seen some very alarming cases over the summer, for example. There is a folk musician called Emily Portman, who had not released a new album but then started getting some of her fan base contacting her about her supposed new material. That was all AI-generated material, so that was very disconcerting for her as an artist and performer. Similarly, there was a band called The Velvet Sundown, which again appeared on Spotify—over a million streams. From a UK Music perspective, we are very in favour of this concept of labelling and making sure that you have identifiers within content, particularly when it is purely AI generated. As part of your inquiry, I would encourage you to think about it. Obviously, this is a discussion between the creative industries and the tech AI developers, but also the digital platforms here, too. From a music perspective, a lot of it will be on Spotify or Deezer and so on, and we have started to have some discussions with them about how we can get to a system where we can have labelling and things like that.

We do think that there is consumer demand for this as well: 83% of people we polled last year, a public, statistically significant poll, supported the idea of some labelling. Back to Isabelle’s point, this is a wider society issue. In context, yes, it is important to have this zonal focus on AI and copyright, but the AI challenge is much wider and we are part of that discussion.

Lord Knight of Weymouth: I hear all that but is a provenance watermark not part of a possible solution for future creation? How does that help us with the very real problem that Isabelle identified, the past infringement? Isabelle, do you have a view?

Isabelle Doran: Yes. Again, there is a challenge with any watermark, for example. We agree with labelling and you will not be surprised to hear that about 96% of our members wanted mandatory labelling, particularly of synthetic outputs. It is not necessarily about the provenance of original works; it is the outputs. Again, the burden should be placed on big tech to label outputs, so those with the services to do that. Essentially, we would require any food manufacturer to place labelling on their food, so why not the same approach there?

On the point about the output and trying to seek remuneration for past works, I think that it is just the way business should be conducted. If you have used or exploited something, you should be paying for what has happened. You should not just be getting away scot free as such. When you look at the considerable amount of investment that big tech and these services are receiving, there is plenty of money there to be able to do that, which is why we are seeing deals being done. They recognise it and the more deals we are seeing, the more that argument fades away about not being able to license for past uses.

Lord Knight of Weymouth: I totally get the principle, but in practice when you have that long tail of small content creators, how do we recognise what they should be remunerated for from the output based on past infringement?

Isabelle Doran: Again, this is where collective rights management has always succeeded, because they deal with complexity. They deal with the complexity of individuals. They deal with being able to identify those works, and together with technology partners they can deliver this. We have been having exactly those discussions with visual arts CMOs. We know that there are good AI developers out there who want to do that and who are interested. There is another company, OpenOrigins, that we have been speaking to, which has said categorically that it can identify works that are out there with the original work. If you combine the fact that we can identify original works in an output and then you can associate other works out there with the original one, there are opportunities. Again, it is about the willingness of those who are exploiting those works in the first place to take part and discuss with CMOs and technology partners to find that solution.

Q12 Lord Knight of Weymouth: Anna and Owen, perhaps you could help me with two final things and then if there are any other brief comments on what we have discussed, they are welcome. The first is distinguishing between AI-assisted and wholly synthetic works. Creators are using AI tools, perfectly understandably. There has to be a tipping point whereby it was a human creation assisted by AI, or a synthetic creation quality assured by humans. How are we able to differentiate? Finally, do you think that we should go down the Danish route of protecting our face, body and voice as personal IP?

Owen Meredith: On the AI assisted versus wholly synthetic argument, from a news perspective I think that it is almost a moot point because where news is unique is that we stand behind everything that we publish. There is the editorial control and legal accountability. It is almost irrelevant how that content has been created if you have that human in the loop and you have the editor willing to stand behind the content. That is not the same for all sectors but it is from our point of view. Similarly, on personality rights and image rights, that is probably not one for us.

On your specific point about human consumption and the question of inspiration and how that works, I do not necessarily accept the premise in relation to AI, but the fundamental difference is that when a human reads a newspaper or a book or listens to a song, they have legal access on agreed terms to that content. At the moment, AI does not even have that legal access established. You might be inspired by something, but if you read a book in a library, you have not paid for it but it has been paid for through public lending rights. If you listen to a music track in a shop, it is through a licensing agreement. They are established practices.

Anna Ganley: I go back to the ingredients of the cake and food standards. You can define what has been used in the creation of that work. We are seeing in contracts a definition about whether something has used AI in an assistive way, or through creative prompts or things like Grammarly. It is just with the wholly synthetic content where we see copycat books being put on online retailers the day after something has happened. You know that nobody has written that book overnight, yet it suddenly appears online. There is no labelling, so for the consumer there is no clarity. They do not know what they are buying—whether it is original, human-authored content or whether it has been AI generated and uploaded, within 24 hours sometimes. The consumer is often left duped and disappointed because they have bought the product, and they have to buy it before they realise what it is. They are losing out and that is why we need clearer standards and clearer labelling of the origin and provenance of all products that consumers are buying.

Q13 The Chair: If we can run on for a few moments, if the panel members are okay, we have a couple of questions. I wanted to ask Owen about the Competition and Markets Authority signalling that Google will be subject to increased oversight and possible conduct requirements. Can you comment on that? Clearly, that seems to be a step towards a fairer digital economy, does it not? I think that is what your organisation said about that. How important is that and what do you think about the timescale of what it is doing?

Owen Meredith: I think that it is extremely important, and this place passed legislation at the end of the last Parliament, which is now beginning to come into force. Google designation was determined last month. The conduct requirements have yet to be put out to consultation. We are probably looking at a six-month timetable before any conduct requirements come into force. As the committee will probably be aware, they will apply to Google only in reference to search, but clearly Google is a market leader in AI, and I think that the CMA can look at where it is using its competitive advantage in search and particularly where it is using its search crawlers to scrape publishers’ content and then using that for AI overviews, for example. Publishers do not have the ability to opt out of AI overviews because they would end up opting out of search as a consequence.

I am optimistic, but I think that this place needs to keep the political pressure and support behind the CMA in order for it to be able to move at pace. Clearly, there is a lot of lobbying coming in the opposite direction, but the will of Parliament has been very clear on this.

Q14 Lord Storey: I want to narrow it down a bit. Right at the beginning, Tom, you surprised me by being quite optimistic. You said we had the capacity to be the “gold standard”. Could you very simply tell us how we reach that gold standard?

Tom Kiehl: I was referring to transparency there. I do not think that there is a country on earth that has this right yet. The EU has a model, for example, around transparency but it is not sufficiently granular. I think that there is an opportunity for us as a country, particularly when we think that the creative industries are worth £125 billion to the economy. It is one of the eight sectors for growth and it is something that we are inherently very good at. We are known across the world for our music output, our film output, our brilliant photography, our brilliant authors. These are things that we are very good at, so when we do not have the country leading the way on the transparency obligations, let us put some effort in and focus on that and getting that right, rather than just trying to strip out copyright. That was really what I was saying.

Lord Storey: Equally, we have the opportunity to go in the opposite direction and get whatever it is—the bronze standard or the brick standard.

Q15 Baroness Fleet: Each of you has explained the problems and the challenges that you have, both from the past and for the future. We are very grateful for your very detailed answers to our questions.

Everyone talks about transparency. They talk about kitemarking and these things. From the point of view of the work of this committee, it would be helpful to get from each of you—maybe you all agree on what the answer is—what specifically you would like us to recommend to the Government to do in terms of regulation. Regulation can cover so many different things. Have you thought about how you narrow down your ask? Is it different for each sector or is it broadly the same?

Tom Kiehl: My message to the committee and my message to the Government is, let us not tear up the copyright framework. Let us focus on transparency and get that right.

Baroness Fleet: In terms of actual regulation, what are you asking for?

Tom Kiehl: From a music perspective, it is to be more granular in terms of the actual mechanisms and particular identifiers, making sure that timestamping where copying has taken place is explicit in its requirements, in compliance with the UK law as a whole.

Isabelle Doran: Of course, I will say the same with regards to transparency. Yes, I would agree with Tom that there should be granularity. We can all contribute to what that should look like and provide much more nuanced information on exactly what it is we mean when we are talking about transparency and traceability, because it does help. To give a brief example, there is a hosting company called Hugging Face that has over half a million datasets; 57,000 of those are image-based datasets, which can contain billions of images. It would be helpful perhaps for us to contribute to a transparency paper so that we could provide that granularity for you.

My second request would be, certainly, a personality right. The reason I ask this is that it is having such an impact on our creators. Anybody can prompt a particular artist. For example, one of our photographers lost a commissioned shoot to gen AI and they were asked, “What prompts would you use in order to distinguish your style?” This goes to the heart of the challenge. It is not just that these machines can output similar works; it is the fact that you can use those works to copy the style of individuals. As I said right at the very start, what creators strive to do is to differentiate themselves in the marketplace. A personality right that sits under trademark, so separate from copyright, expanding on passing off, which falls under trademarks, would be particularly helpful. There are good examples around. Lord Knight mentioned Denmark, but obviously that is to do with an individual person. What I would be interested to see is protecting the style of our creators, so that it is distinguishable in that way.

Lord Knight of Weymouth: “In the style of” has never been within copyright law to date.

Isabelle Doran: No, exactly, but it can fall under trademark, under passing off.

Lord Knight of Weymouth: Yes. It creates problems for satire and a bunch of other things.

Isabelle Doran: Exactly, but you could have an exception for news reporting and personal uses and so on, so it is possible.

Q16 Baroness Fleet: Anna, do you have very specific requests for us to bear in mind, and that you would like us to recommend?

Anna Ganley: Yes, I have five requests: a comprehensive framework of transparency requirements, which we think should include mandatory disclosure of specific materials used to train AI models; auditable records of how personal data and copyright-protected data have been collected and used; transparency around how AI systems operate and make decisions; clear labelling of AI-generated outputs; and international consistency, because this is a global technology, so that will help to prevent any regulatory fragmentation.

Owen Meredith: The first request is a clear statement from this committee supporting the existing UK copyright regime and backing something similar to what we have seen in Australia recently, completely ruling out a text and data mining exception. The second request is transparency regulations, and I would encourage the committee to take a look at those proposed by Baroness Kidron in the passage of the Data (Use and Access) Bill as a good starting point. They need to be both flexible and future proof. You do not want granularity in primary legislation. Thirdly, in order to provide clarity and speed in resolving these issues, we need a clear statement from the Government that any AI developer wishing to operate and deploy its products in the UK must negotiate with creative rights holders, and a clear statement that if that is not followed, there will be legislative reinforcement.

Baroness Fleet: Thank you. That is very helpful and very specific.

The Chair: I am still getting requests from members but I think that we ought to stop there. We have run over.

Owen Meredith: I am happy to follow up in writing if there is anything that members wish to—

The Chair: That would be helpful, thank you, if any of you have thoughts you want to send in. You will receive a copy of the transcript, so if anything in it does not seem to reflect what you said, you can correct it. Thank you very much for your time with us today.