HERD, Ms Annabelle, Chief Executive Officer, Australian Recording Industry Association; and Phonographic Performance Company of Australia KENNEDY, Mr Simon, President, Australian Association of Voice Actors LIM, Ms Teresa, Vice President, Australian Association of Voice Actors PULLEN, Ms Claire, Group Chief Executive Officer, Australian Writers' Guild; and Authorship Collecting Society Australia Aotearoa [by video link] RAMSDALE, Ms Cecelia, Secretary, Australian Association of Voice Actors [10:47] CHAIR: Welcome. I understand that information on parliamentary privilege and the protection of witnesses giving evidence has been provided to you. Now is an opportunity to provide some opening statements, but we'd appreciate it if you can keep these relatively short. As you can see, there are a lot of questions that committee members are keen to ask. If you want to, please make a very short statement, and then we'll be able to take any further written information on the record via the secretariat. Ms Herd : Thank you for the opportunity to appear today. I'm here representing Australia's recording artists, international recording artists, and local and international major and independent record labels and distributors, representing US$31.7 billion in global recorded revenue. Record labels globally invest about US$8.1 billion in making and promoting music. Elon Musk, Dario Amodei, Sam Altman and Bill Gates have two things in common. They're the most powerful AI company leaders in the world, and they've all warned that the technology they're responsible for poses an existential threat to humanity. Last week, the Prime Minister said there were dozens of cases of AI agents accessing information without authorisation, including US government sites and our beloved Medicare. How do we manage this risk when no-one, not even the AI leaders themselves, understands where this technology is taking the world? OpenAI and Anthropic want the Australian government to sign up to a deal to train their frontier AI agents here on the proviso that they don't have to abide with our copyright laws. Their reason? The world's richest and most advanced companies say it would be too costly and complex to work with rights holders. It would also harm the numerous copyright infringement court cases they are fighting in the US, where they admit to copying music, films, books and media sites without permission to train their models. Their plan is to give our artists and record labels some token amount of money to forgo their property and creative rights, and those who don't want to play will have to find a way to digitally protect their works from the AI internet scrapers that these companies can't even control. In other words, Australia's artists will be the roadkill in the rush to this AI deal. In July, the Prime Minister promised that artists must retain control of the price and value of their work; anything less is theft. No amount of spin will convince Australia's artists and record labels that what big AI is proposing is anything short of theft. And I cannot think of a single thing that would be more damaging to the future of the music industry and the broader media and creative industry sector than caving in to big AI. ARIA and PPCA are not against these companies doing their frontier training in Australia. We understand why that is beneficial. But it should happen on our terms, not theirs. There is nothing less sovereign than rewriting our laws to suit foreign company commercial interests. Done the way that big AI is proposing, it will devastate the music industry and make Australia a pariah for rights holders globally. And this is also unnecessary. Around the world, major AI and tech companies like Spotify, Udio and ElevenLabs are entering into licensing arrangements with rights holders now, and music is easy to license at scale. We have efficient and sophisticated global frameworks that license music across all digital platforms around the world. You can license 90 per cent of the world's sound recordings with four global deals, one with each major label and one with Merlin, which represents independent labels. Even the smallest self-releasing artists in Australia license their music to Spotify and TikTok within this framework. But we are being asked to believe it is too hard for the world's smartest and richest companies. There is clearly no market failure—just two foreign companies, each valued at over $1 trillion, asking government for a free ride. Thousands of small businesses and millions of Australians who work and participate in our creative industries are relying on our elected representatives not to cave. If you support culture and intellectual endeavour, if you believe in the rule of law, if you subscribe to the economic principles of functioning commercial markets, reward for effort and government intervention only when it is necessary and justified, you will not support the proposal on the table. CHAIR: Be as short as you can. Thank you. Mr Kennedy : AAVA represents the voices of Australia—the voice artists, the voice overs, the people who make what you hear on radio or watch on screens sound Australian. And these are people who don't own copyright. We're here to make it very clear that AI has a transparency problem and a massive consent problem. Any child or teenager will tell you that consent means you ask first. If the answer is no, no means no, especially when it comes to their bodies. And, in an AI context, that means your voice, your face, your physical image and your likeness—your identity. Now, the men who run AI companies have been acting as if consent is assumed, and they are using their millions of lobbying dollars to rewrite the rules of consent in Australia. When a voiceover artist reads a commercial for a company, they consent to getting paid for one specific job. They do not consent to an AI platform scraping their biometric data to create a large language model that can better impersonate a human being. When parents consent to their child's photograph being used for the school newsletter, that alone is what they consent to. They do not consent to an AI company scraping their child's image so strangers can make a more convincing deepfake of their kid. Opting out after the fact is too late, because you cannot un-violate somebody. Biometric data is not just zeros and ones; it is the identity of every Australian, and Australians deserve protection. They also deserve and need transparency. Right now, AI generated voices and faces are pretending to be real people and are convincing Australians to put their savings with Bankwest, place a bet with Ladbrokes, trust this news bulletin and maybe vote for this party, not that one. Mandatory labelling of AI generated content should and must be legislated. Transparency is not too much to ask. Tell us in real time when that AI generated content is AI generated content, and prove that the data that made it possible was obtained with informed and absolute consent. CHAIR: Thank you. And online? Ms Pullen : Warami and kia ora from Auckland. The Australian Writers Guild represents over 3,000 performance writers in screen, theatre, podcasting, comedy and audio drama. The Authorship Collecting Society Australia Aotearoa is our collecting society—our collective management organisation. All AI large language models have been built on work stolen from my members and members of other unions and societies. The late, great ACSAA member, nationally beloved satirist John Clarke, said, 'In the scientific world, when they see that something is happening again and again and again, repeatedly, they don't call that old hat. They call it a pattern.' Knowing the pattern from these tech platforms, we have to behave like we know it. Our PM agrees with John, as far as I can tell. Referring to the world-leading social media ban that Australia brought in, he said: 'Imagine if the world had acted a decade ago. Imagine the difference it would have made if these limits had been put in place when the world first grasped the risks of these platforms.' These same platforms that the PM talked about are the AI companies that want our copyright laws weakened to legitimise theft after the fact. Big tech has a pattern, and we have all seen what that pattern has meant for us as Australian citizens, with the hacks of Medicare and our other institutions. Like my creative industry colleagues for years now, we have been saying to government and to big tech, let's talk. There is a way to work out something that satisfies everyone's needs. We are yet to get a response from tech. So, over those years, the government has worked, creators have met, and big tech has paid lobbyists. There is a solution, not just in copyright but in transparency over what has gone into the models, not just for creators but for all of us as citizens, ensuring that creators have meaningful control, that, when they say yes, they are paid properly, that we respect Indigenous, cultural and intellectual property and, importantly, that creators can say no. Fewer than four per cent of Australians have trust in AI. Australians know and care that we have all been stolen from and that these companies are treating us like mugs. We can do what the PM said in his speech in July: 'Our actions can send a signal and set a standard.' That standard could protect my members and all Australians, and we remain ready to help design it. It is not creators who are the problem. CHAIR: Excellent. Thank you very much. I think that is a really good way to talk to the questions that I have. What we're hearing at the moment is that AI training—offshore, not here in Australia—is resulting in a lack of transparency and resulting in creators and rights holders not having control or being compensated. And that is what we are seeking to find. I've got two questions on that. One: what do we need to get a hold of in terms of these AI developers to determine whether their works have been used in training? What does that transparency really look like? How do we secure that? And right now it would seem that the status quo—what is happening now—is enabling companies to train without us having a say or even knowing about it overseas. Can I get a sense from the panel here today what your views are in terms of the current status quo working? How do we design or use this opportunity to potentially get something better for creators and rights holders? Ms Herd : This concept that the status quo isn't working is being used as an argument by AI companies to change our copyright law. You don't address a systemic problem of mass piracy by changing the law to legalise that piracy on an ongoing basis. Most of the training of these particular LLMs that are the focus of these sessions, from Anthropic and OpenAI, as far as I can tell, has happened in the US, and there are more than 130 copyright infringement cases currently underway in the US. That suggests that a lot of rights holders think that piracy has occurred in the US. In Australia we have a very clear system, where, if you can't rely on a fair dealing exception, you can't copy. The law is very certain. It is very simple for these companies to come to us and say, 'We'd like to license your content.' They should come to the negotiating table, not to the Prime Minister's office. Ms Pullen : If I can jump in—on that question, I agree with Annabelle. What I would say is, whether or not the status quo works depends on what you think the problem is. There is always going to be a problem in any system that gets designed with any individual, person or organisation that is prepared to flagrantly break the law. That's just the issue here. I observed some of the earlier session where it was raised that the training that's happening in the US has all been lawful, and that's simply not the case. There have been very large settlements paid out in the US around how some of the conduct has been unlawful. So I don't think the idea that the status quo is better overseas and somehow we need to follow that here is quite made out. I'm happy to speak to the transparency question further in a submission, but there are a number of territories in which the AI companies are required to be transparent about the data that they have ingested, and I think the Californian law chapter A17 is really instructive on this. It represents no additional burden to these companies, of course, because the majority of them are based in California. In looking at other places in the world and how they've done well on the transparency piece, I would suggest we start with California. This is something that is not necessarily a question of copyright but, again, speaks to all of us as citizens. Our data is in there. Mr Kennedy : When it comes to the idea of transparency, what we're dealing with is data. Data is probably one of the easiest things to catalogue and track in the world. These people who create these systems are all about data. They know what goes in. They put it in there. It's zeros and ones, and there's nothing harder to track than zeros and ones. So the idea that it's too hard is rubbish. I also think that the status quo not working is a little bit of a furphy, because it's akin to the police officers standing by, watching everyone loot the shop and saying, 'Well, the law's not working.' How about we enforce the law and not try to change it on behalf of the pirates to make it a little easier to work with the pirates? I just find that that is a distraction from AI, and I think that it's not the basis on which the argument should be formed. Ms Herd : It's really important as well to note that this line that no-one's getting paid right now is actually not correct. I've provided as part of my submission a long list of licensing deals that have already happened. That market is growing. Those licensing deals are between major record labels, independent record labels and AI companies. It's early days in this market. It is developing, but it is growing really rapidly. Those deals will ensure that royalties will flow to recording artists and record labels here and overseas, because music licensing operates in a global framework. You don't do different licensing deals for different territories; you do a global deal with TikTok or Spotify or YouTube Music or whoever, and that deal enables copying and making available this music in almost every territory around the world. So those deals are already happening, and to pretend that that licensing market isn't there is wrong. Ms Pullen : It does also pretend that payment is the only issue. From our surveys of our members, the majority of them would prefer that their works not be used and that they are therefore not paid. They would prefer to maintain control over their stories. They would rather not be paid. The idea that payment is the only thing we're haggling about, I think, is a step in the wrong direction before we go back to what these AI companies have actually taken and whether or not artists have decided that they want them to have it. Mr Kennedy : I double down on that on behalf of the performers. Most of us would rather not be involved and not have our biometric data involved in a large language model. I would rather not get a cent from AI and keep working the way that I am working right now in giving my voice, and I'm sure that my colleagues would agree with that. It's not about payment; it's about consent. Ms Herd : My industry record labels have been more than open to talk to AI companies about doing deals and enabling this very exciting and innovative technology, but the ability to control how those tools use music is critical because the existential threat to music is not just about the money; it's about the substitution of our real music with AI music. And that's already happening. Sony has taken 200,000 fake AI tracks down from DSPs. This kind of music would replace real music in places where we earn royalties—for example, in bars and clubs as background music, in advertising or wherever. It's not just about the money. In fact, no amount of money can really replace those royalties that are not just about AI royalties. They're about royalties across all of our existing revenue streams. CHAIR: Are you suggesting that there isn't a way in which we can better improve the current framework to improve control as well as potential compensation? Ms Herd : Well, I think the thing that's hindering licensing in this market is this exact conversation and the fact that so many people in Parliament House are willing to listen to these companies say: 'It's too hard to licence. Can you please change the Copyright Act so we don't have to bother in Australia?' CHAIR: I'm also hearing from creators themselves saying they're not getting— Ms Herd : I think there are things you can do. I mean, we've talked about transparency and better means to enforce our rights, but removing those rights is not the way to deal with it. Mr Kennedy : If you built a car with stolen parts, you wouldn't get to keep the car. CHAIR: I appreciate that. I don't think anyone's looking to make that analogy. Mr Hastie. Mr HASTIE: I appreciate your concerns around transparency and consent. Have you had any engagement with the government itself about the way this will be done? Have you had a chance to speak directly with— Ms Lim : We believe that mandatory AI labelling at time of broadcast is really important. The issue here is that the tech has advanced so rapidly over the past two years that people can clone and create replicas, and there's material out there online, on TV and on radio where you and I don't know if it's human or AI. We're saying that all Australians deserve to know if what they're consuming is human or AI so they have the choice. It's like food packaging labelling. We have that in place so that we can read the ingredients on the back of food labels and go, 'I choose to consume this as opposed to this.' We're saying that all Australians deserve to know what they're consuming and to be able choose whether they want AI or human. An example that we came across last year was with regard to radio. We'll come back to radio; that was brought up just before. The Australian Radio Network started broadcasting an AI avatar over six months and presenting her as a real human without disclosure that she was actually an AI clone. It was discovered through the work of a journalist, Stephanie Coombes, that they were just doing text to speech. There are lots of problems with this, but this was an Asian radio host that they put forward—a female Asian radio host that was the only female Asian radio host across their entire network—and it wasn't actually her. It was four blokes doing text to speech. People listened to this station, tuned in and chose to consume that without realising that she wasn't real. We see a massive problem with that. This could have been solved with an easy disclosure at time of broadcast saying, 'This show is AI generated.' We believe it's really important that there be adequate labelling and we don't just let people label as they please, because there's going to be insufficient disclosure that way. We believe that this is where we need to step in and have sufficient disclosure. So we approached CRA, the Commercial Radio Association, last year to say, 'Hey guys, with the new code of practice, we believe that AI should be included.' So they included it. They've included section 11, pertaining to AI transparency. This is on page 23 of 40 of our submission. But what's been added now allows broadcasters to use AI generated audio with very little disclosure and no requirement for disclosure on the medium the content is heard on. This means we could have AI generated content on radios and then the disclosure in a little bit on a website that no-one's ever going to read, and we think that this is not sufficient as all Australians deserve to know that this program, this music, this announcer, this everything is AI generated. That needs to be really clear at time of broadcast. Mr Kennedy : You asked about the governance and where that stands. Media is a self-regulating industry. Ms Lim : And self-regulation isn't working in this. Mr HASTIE: What about personal contact? We've got the committee here. This is about setting the framework or the foundation for a settlement around AI that all Australians can buy into and are agreed upon. But have you had direct contact with the Australian government? Mr Kennedy : Absolutely. Over the last few years we've probably had— Ms Herd : about five trips to Canberra? Mr Kennedy : Yes. Mr HASTIE: Do you feel like you're being heard directly? Ms Herd : Recently in August, we had 10 meetings in one day, and it has been the best day we've had so far because the penny has dropped that it's time to actually move and move quickly on making sure the regulations are in place to protect all Australians' voices. Transparency is a big part of that. Mr Kennedy : I would say that, generally, our parliamentary representatives are very good at a nodding. I don't mean that in a cynical way. I mean that they hear what we say and they're like, 'That makes plenty of sense.' Ms Ramsdale : The solution is harder. Mr Kennedy : Of course, then action is required. We do need to move. We have been speaking about this, as most of my colleagues have, probably for a few years already. Our members are already 62 per cent down on their work in the last 12 months, so things are happening. But self-regulation is not working. Ms Ramsdale : Also, as creatives at the coalface, we have a lot of conversations with people in our lives who are not creatives, and it is amazing how often people will be engaging with AI systems and not thinking about the greater ethical or moral situation that they're engaging with. I think, as a country, we actually have an opportunity to have some education around the use of AI systems and, as individuals, what responsibility we have. I always come back to—do you remember the 'you wouldn't steal a car' commercial that used to be on DVDs and VHS. Mr HASTIE: With the music. Ms Ramsdale : Yes. Everyone remembers the soundtrack. I think perhaps we need something like that, which tells the public, 'Great, you get to make a funny picture on ChatGPT of your aunty, and you've taken the photo, but did you ask your aunty whether she wanted her photo added, because now that data is part of the ChatGPT universe? Ms Herd : I think all of these kinds of things about this proposal that's on the table, it's basically saying that all information that Australians put on the internet, or that anybody in the world puts on the internet, is openly available, and is suddenly allowed to be trained on in Australia unless those people put digital protections around their content at the asset level everywhere in the world. Ms Lim : Yes. And there are four pillars that we always talk about: consent, control, compensation and transparency. We believe that those need to be in place in order for AI to be ethically used. And absolutely what Cecilia said: tech has scaled so much, but ethics has not. We've stayed down here with ethics, and people need to be educated. Mr HASTIE: One last question, just so I fully understand your position. If we zoom out and look at the big picture, given that AI is here and it's not going away, do you think Australia is better off having a sovereign capability, even if that means inviting frontier labs here that operate under our laws rather than us being tethered to a submarine cable all the way to the United States and having very little leverage if there are copyright breaches in the future. I guess that's the fundamental question. Then, that being the case, if you agree with that, it's just a question of working out a settlement that works and protects Australian creatives. Ms Herd : I think we absolutely understand why it's important and why the government and this committee are considering how we encourage the best access we can possibly get for Australia to this technology, to the frontier models, to everything else. I believe Sam Altman last week said very clearly that, whether or not they do training in Australia, they will provide access to those models. My question is: why does Taylor Swift's, the Beatles' or Jimmy Barnes's music need to be thrown under the bus in order for our government to regulate companies that are operating in our jurisdiction? If it's critical infrastructure, and there are many other examples of companies that are deemed to be critical infrastructure, we regulate those companies because they operate in Australia. These companies have offices here. They're providing products here. What is the magic thing about training in Australia that's going to allow us to regulate it? Is it just because that's when they'll agree to be regulated? I'm not sure. The other thing is: what is the problem that we're trying to solve here? What is the access to the content that they need? We're talking about the biggest change to the Copyright Act since it was brought into existence, including changes that were made when the internet was invented. Taking away the rights of copyright holders—you would not do that unless you had really strong evidence of a necessary public intervention. The only evidence I can see is companies just stating that it's too hard to licence. Can you show me one example where they've approached a rights holder and they've been told no, or it's been too difficult? Can you show me one piece of content that they've tried to licence that they haven't been able to licence? Do they need access to all the content in the world? Are they putting all of the content in the world into their content models right now, including illegal material or other material? Are they filtering that out? There are so many questions. Where's the economic modelling of the impact of all of this on our industries, on the media industry, on First Nations communities? We're rushing into something without even understanding the basics of what we're talking about. Mr Kennedy : Apart from the emotional and the ethical, the economic impact is huge. There's a TV commercial that's been running for the last year for the telco Amaysim. You might have seen the ad. It's a very orange ad. It's AI generated footage. Amaysim is a company owned by Singtel, not an Australian company. The whole ad was AI generated—ironically, I think, bar the voice. I come from a family of actors, and I know how many crew, how many support staff, how many actors, how many people work on a TV commercial to get it made over a few days. These are hundreds of people and hundreds of livelihoods, mortgages, rents and food on the table all paid through the advertising industry, which is a huge support for the creative arts in Australia. Advertising is where the bread and butter is made so that you can do Priscilla, Queen of the Desert. That one ad was visually AI generated, which meant that hundreds of Australian jobs and cafes local to where they shot it—none of them saw a cent. None of them. Australian tax dollars—nothing, because there were no people working, no jobs, nothing being sold. That money was all absorbed, or I should say saved, by Singtel. The only beneficiary would have been the AI company whose credits were purchased, who were based in Silicon Valley. So all of that money that would have been spent in the Australian economy as part of our creative sector, which happens on a regular basis, left our shores. Gone. Ms Pullen : On the sovereignty question, can I just say that to have a truly sovereign capacity, we would build it here in Australia on Australian servers, in Australian data centres. Leaving that aside, I have asked in meetings with government to explain why it is that these foreign firms, to have capacity in Australia, need to steal from David Williamson and Leah Purcell, and there's not an answer that I've received to that question. To be fair, I'm talking to government about that rather than the actual AI companies. One of the questions I also asked was: if we are building a model here that is critical to our national defence, fantastic. My members would probably feel differently about that and how their work is used if that is a black box that the works go into, and if there's no commercial or consumer product generated from it that is then sold back to us and that puts my members out of jobs. I think that's the distinction that I would have loved to be able to talk to the tech companies about, explaining that relationship. There is no need to steal from Australian playwrights and screenwriters to build sovereign capacity that I can see. These companies will be making commercial products available here built on stolen work. So I suspect my members would be very open to the idea of some kind of settlement where it relates to sovereign capacity, but the link between that and these commercial products now that are built on theft is unexplained. Mr HASTIE: That's very helpful. CHAIR: I'm conscious of time. I know Senator Pocock has a couple of questions, and there are a couple of other members. Over to you, Senator Pocock. Senator DAVID POCOCK: Senator Cadell has asked the voice actors to respond to his questions to the previous panel on notice, if you're happy to look at the transcript and do that. Ms Herd, there's been talk of some kind of levy or AI companies paying into a fund because it's better than nothing. What percentage of disbursements from that fund would go to Australians versus internationals under copyright law? Ms Herd : I think this is a really important point that seems to be missed in the discussion that I've seen. Under our international copyright treaties, you can't actually take action that removes copyright protection here and only compensate Australian rights holders. It's a concept called national treatment. It basically means that we provide reciprocal protection for foreigners' copyright material in Australia, and they do the same thing. In terms of the percentage that would go overseas, it would be really hard to work out. Even in music, if you look at music, unfortunately streaming of Australian music is at a very low number at the moment; it's under 10 per cent. So, for recordings, you would say that under 10 per cent of that money would go to Australian rights holders. Senator DAVID POCOCK: I've heard a quantum discussed: $300 million. Noting your opposition to this model, is $30 million going to all Australian creatives, if the music example holds, fair compensation? Ms Herd : You'd have to split that. Recorded music revenue in Australia alone was over $700 million last year. You're taking $300 million and you're splitting it between every category of copyright owner—commercials, film, architects. Everyone who uploads anything online could be described as a copyright owner—globally and in Australia. So the percentage staying in Australia would be minuscule, and the amount distributed to any one person under that scheme would be minuscule. Mr Kennedy : It's also hard to determine, without transparency, what data there is. How do you really know? Ms Herd : And it's impossible to differentiate the values of a poem somebody's uploaded from their school project and a Beatles recording. Are we just going to lump them all into the same amount and say, 'You all get the same amount'? Everybody in that mix is entitled to their payment out of that amount of money, so it's not going to go far—let alone if there's an admin fee taken out of that, because the government is going to have to administer all of this, which will be a huge job. Ms Pullen : No, they aren't, because there are things called collecting societies, and we administer these regularly. To the question around the adequacy of the $300 million: we're a member of our international society of collecting societies, and the annual figure that CISAC members collect is in the order of 14 billion euros. The $300 million doesn't come anywhere near even touching the sides, I would suggest. But we also need to go back a step and ask what you are paying for. Is this an annual fee that we're somehow meant to split up, or is this a one-off for the overseas theft that has already taken place? I think, before we talk about how much the money would be, we should talk about what it's for. Something that I have heard relayed to me is that the tech companies say they don't want my members' work. I don't think that's true, because they've stolen it. But, if we had a transparency regime in place and my members' work wasn't in it, then we would go away and there would be nothing to argue about. So I think, before we talk about what the amount of money would be and who would split it up and how much would go overseas, we first need to talk about what it's for. Is it to deal with the fact that they have stolen from us, or is it an ongoing payment for ongoing use? Senator DAVID POCOCK: There's this interesting tension here, where we're hearing that there's fair use in the US, so the status quo is that you all get nothing, and something's better than nothing. But we know that there's a whole bunch of lawsuits in the US, and probably the most high-profile one at the moment is the New York Times lawsuit against OpenAI and Microsoft. In documents revealed in that, Microsoft director Brent Hecht is quoted as calling the AI build-up 'the largest theft of labour in human history'. He has privately stated that applying a fair-use defence to AI data scraping would 'make a complete mockery' of the concept of fair use. So, given these are working their way through the courts in the US and may take some time, what's at stake here if Australia rushes ahead and creates some sort of workaround? Ms Herd : To me, this seems to be a deliberate strategy by the AI companies. They're open in saying they won't license here, because it could impact their US lawsuits, but the converse is also true. If we go down a path and declare that there's no licensing market—'Australia is deciding to upend the global licensing market and say there isn't one'—then that will assist them in their US lawsuits, and I'm not sure why that's in Australia's interest. Also, there was a case last week, from one of the US appeals courts, which was Reuters versus an AI company. It found that the training for the AI tool was not fair use, because of the fourth factor in fair use, which is substitutability in the market and the impact of the product in the market for the existing work. And that case, which was an appeal case, found that it was not fair use, because it did impact the existing market for the work. And there are many cases in the US that say that that is the most important fair-use factor. The people I represent are suing at least two AI companies in the US—Anthropic is being sued by music publishers—and those factors are very much going to weigh in those cases, and it is very much undetermined, at best. Mr Kennedy : I think that section 43B of the Copyright Act is supposed to cover that, but I would say that this is not a temporary copying situation. This is literally a stripping down for parts, reassembling them in another form and putting them back out into the marketplace. Once again, I will reinforce that people's faces and their voices and their identities aren't covered under any of this. So we are left wide open, in that respect. CHAIR: Thanks, Mr Kennedy. Senator Sheldon. Senator SHELDON: To the Writers' Guild: Ms Pullen, you mentioned that the copyright laws are largely sufficient but there are areas that could be improved for rights holders—greater consent, control and compensation—when it comes to AI. What are some of the key design features the new model would have? Could you briefly take us through that. Ms Pullen : One of the things the actions of these firms has made clear is that we need to improve the copyright frameworks around both enforceability and who has the right to do what. I'm usually reluctant to recommend that we shape laws around the worst offenders, but in this case I think it's probably warranted, particularly given the events around Medicare. One of the things we could do is strengthen the economic and moral rights of creators to make it clear that they have both the right to be paid when their works are used in training and the right against having their work used in training. One of the things that AI is novel in is that it can reproduce people's voice, face and likeness. Those things are currently not protected by copyright. As AAVA and MEAA have quite rightly point out, this is a new problem. Frankly, it's one that I'm particularly concerned about as a citizen, in terms of what political advertising is going to look like. You could create a video with an MP eating a puppy, and it would look exactly like them and be badged as such. There is no requirement to tell people that that's what they're watching—that it's an AI generated fake. I think some of the states that have started criminalising AI deepfakes are a good model to follow. The other thing that we probably need to talk about is First Nations works. I don't see any way to allow the ingestion of these works into models that is appropriate and culturally respectful. I think it's probably worth pressing pause on what we may or may not do in terms of our copyright framework in that case. As I said, in general, the copyright framework works. The problem is that we have repeated bad actors who have said they're not going to respect it as it is. Clarifying that creators have a certain set of rights—both to be paid when they authorise their work and a right against having their work taken against their will—is probably where we need to start. CHAIR: Thank you. I'll leave that as the final contribution for this session. I appreciate you all taking the opportunity to come before us and answer our questions. We'll now suspend briefly while we get ready for our next panel. Proceedings suspended from 11:26 to 11:37 .