0:01 I'm Daniel James, and you're listening to 7am. Anthony Albanese has made his love of Australian music a big part of his public identity. But now some of the Prime Minister's favourite artists, including Jimmy Barnes, Paul Kelly and Peter Garrett, 0:19 are warning him not to give in to the world's biggest tech companies. The musicians fear the government could weaken Australia's copyright laws, allowing companies to use their work to train AI models without permission or payment. And while the 0:33 Prime Minister insists artists should retain control of their work and be paid for its use, he hasn't ruled out changes to the law. Today, we revisit a conversation with copyright lawyer Hannah Marshall about the battle between Australia's creative 0:46 industries and the tech giants and why billions of dollars could be at stake. It's Sunday, November 15. This episode originally aired in August 2025. 0:58 Hannah, let's begin our conversation about AI and the creative industries with what we're starting to see in terms of the broader impact on anyone who makes a living making any kind of art in Australia. 1:13 The creative industries in Australia are on the cusp of crisis and we're seeing increasing outcry from people authors, musicians and representatives of those industries about the fact that these AI companies are basically coming and using their content, deriving 1:30 commercial value from that and not paying anything back to these creatives. So they're not seeing any upside, even though their works are being used. 1:40 Nearly 11,000 creatives put their name to a letter that described the unauthorised use of their work by AI as... 1:48 and I quote, major and an unjust threat. 1:52 We're seeing groups like Creative Australia come out and say that they want to have principles for using their content in generative AI framed around fairness and the right to be remunerated. And we've also got prominent authors like Richard 2:07 Flanagan calling this the biggest act of copyright theft in history. 2:11 His books, quite interestingly, were actually part of the trove of books which were the subject of a case against Meta in the US about whether they had infringed copyright in those books by using them to train their AI model. 2:25 They accused Meta of illegally using their copyrighted works to train its flagship generative AI model, Lama. But instead of the declaration that Meta violated copyright law, the judge ruled the plaintiff's arguments weren't strong enough to move forward. 2:41 And let's just spell out what is actually happening here. 2:44 So we're talking about AI companies basically scouring works of literature, articles, news, and using that material to train models to create better AI. 2:54 Is that right? 2:55 So when the AI model ingests content, there's an underlying technical question about whether it's making an actual copy of the copyright content in the learning process or Everyone tends to assume that it is, and none of the AI 3:12 companies are kind of loudly coming out and saying that it doesn't. But that would be the precursor to the copyright infringement happening. If it's not making a copy of it or reproducing a substantial part of the book or the song or whatever it is it's learning from, then 3:26 there wouldn't be a copyright infringement. But we're not seeing that argument being seriously thrown out by the AI companies. 3:33 And so for the most part, people are assuming that it must be some form of reproduction. And therefore... there's a potential for infringement absent some defence applying. 3:43 And that, I suppose, brings me to my next question. 3:46 Is this breaking the law, what these companies are doing, or is this a grey area? 3:53 Look, if it's reproducing it, then that's a copyright infringement unless a defence applies. And so that's kind of the bottom line. So under Australian law, we've got quite a narrow set of defences. And so it's not clear that 4:06 there is an obvious defence that the AI companies could rely on, other than saying there was no infringement in the first place. 4:14 So if that is the case, tell me a bit more then about how the AI companies are currently justifying what they're doing. 4:22 So last week, Scott Farquhar, the co-founder of Atlassian, came out quite loudly saying that copyright laws in Australia are holding back the growth of an AI industry here and that we need to change the laws urgently to facilitate the growth of the industries. 4:41 We remain an outlier when it comes to copyright. So this is a barrier to AI companies who want to train or host their models in Australia. 4:49 Specifically what he wants... is a new exception to our copyright laws for text and data mining. 4:55 So my first ask today of government is for the Attorney-General to urgently amend our Copyright Act to look at fair use and text and data mining exceptions. Fixing this one thing could unlock billions of dollars of foreign investment in Australia. 5:11 And just as Scott Farquhar, who is a billionaire, called for looser rules... We had the Productivity Commission release a paper arguing for the same thing. So what is their case in favour of this? 5:25 The Productivity Commission were very loud and clear on the money to be made in short and the jobs to be gained from a growing AI industry in Australia. They said it could boost productivity by between 0.5% and 13% over the next decade, which could be hundreds of billions 5:44 of dollars. So it's a lot. 5:46 Artificial intelligence is a game changer for our economy. It has the potential to support all of our major economic goals, making our economy more productive, stronger, lifting living standards over time. 6:01 And Chalmers came out last week as well, talking about how artificial intelligence will completely transform our economy. And he said that the government's intention is to regulate as much as is necessary to protect Australians. but as little as possible to encourage innovation. 6:20 The only way to make our people and workers and industries beneficiaries is if we treat AI as an enabler, not an enemy, of what we want to see in our economy. That means listening to people. 6:34 But it's interesting that when other governments are exploring and reporting on AI, for the most part they're focusing on how do we minimise harm, from the AI industry. Whereas this report, it's really talking about how do we take 6:47 away the regulatory hurdles so that it's easier for the AI industry to come here and use our creatives content. 6:56 I mean, there's this really strong common thread about the desire to unlock the benefits of the AI industry. And, you know, the creative industries are understandably really concerned that they'll be railroaded in all of this excitement about AI 7:11 jobs and money. And looking at it globally, what you can see is that it's really hard to strike a good balance between the opportunity of the industry, which is undeniable, but also protecting creative industries. And the risk is that 7:27 because it's hard to strike that balance, that it will be done badly or not at all. 7:35 Coming up, authors versus the tech giants battle it out in court. Hannah, let's talk about how this is playing out in other countries, specifically the US, where a lot of these AI companies are based. What are we seeing happen in court? 7:57 So over in the US, they have a broad exception to copyright infringement for fair use. And that's been tested already in cases against meta and anthropic in relation to books. 8:10 And it's the cases that I mentioned before, which included Richard Flanagan's works, And the issue there was that a trove of 185,000 books had been used to train the AI models and, you know, from the author's perspective, that 8:25 was a rampant example of copyright infringement. Applying the fair use defence in the US, the court ultimately decided that it wasn't a copyright infringement and the reason was that it wouldn't impact the downstream sale of of those books. 8:45 So the AI models were ingesting the book content, but they weren't spitting it out again. So you couldn't go on to say chat GPT and say, give me a copy of Richard Flanagan's latest book. It would say no. 8:59 So because the training wasn't being used to create a copy of the work, the court's decision was that it wasn't an unfair use of their work. And as such, the the copyright infringement claims fail. 9:16 Okay, so if we. 9:18 Are to accept the argument that AI might be using this intellectual property, making money off it, but that doesn't necessarily impact someone still going ahead and buying it for themselves, how does that apply when we think about non-fiction, when 9:35 we think about journalism and news? 9:38 Yeah, I think the analysis starts to change significantly when you look at nonfiction, but particularly news, because the value of those works is largely in the information they're conveying rather than necessarily the specific words used to do that. 9:55 And so that means that, you know, using that US legal analysis in the books case, you might get a different outcome because if AI goes and ingests all of the 7am podcast library, all of the Sydney Morning Herald, it can spit out competing content and people could quite 10:16 naturally use the AI tools to access and get news or non-fiction information from them in a way that does impact the market for the news publishers. And so that's the reason that I think AI is presenting almost an existential threat to the news industry. 10:35 So Hannah, if you wanted to both chase the economic opportunity of AI while protecting the creative industries, what do you think a good model that could do both would look like? 10:48 In terms of legal protections, at the moment in Australia, the current copyright laws do seem to create a degree of protection for creatives. And that's what Farquhar's arguing against, is that there does seem to be a standing assumption 11:02 that there is an infringement, in which case the AI companies have to buy a license. You know, that still seems like a reasonable outcome. If we were to go further, though, and create a new law which mandates payment, that might 11:17 add additional protection for the creative industry, but you would need to be really careful about unintended consequences. And we've seen that play out before with well-intentioned legislation in the form of the News Media Bargaining Code because deals which 11:33 facilitated payment by digital platforms through to the news companies were not transparent. They didn't cover the whole industry. They weren't renewed. And ultimately, to avoid being caught by the law, Meta exited the news industry in Australia altogether, which was 11:49 a long way from the desired outcome of the law. 11:54 People like Scott Farquhar are saying Australia's copyright laws are a barrier to growth of the AI industry, but that hasn't really been examined in detail. I can't Imagine, given the funding and the revenue of these massive AI companies, 12:11 that the cost of licensing the content that they're ingesting would be crippling or prohibitive. Nobody's really spelled out in detail why the AI companies can't just pay for a licence. 12:24 And that's the thing that really strikes me about all of this. 12:26 Hannah, thank you so much for your time. 12:30 Thank you for having me. 12:38 Tomorrow on 7am, Pat Cummins is adamant. He didn't know sandpaper was being used to tamper with the cricket ball in South Africa in 2018. But retired Australian batter Usman Khawaja has a very different account. So does Australia need 12:53 to reckon with sandpaper gate again? And who's telling the truth about the sporting scandal that just won't go away? 13:01 I think you can believe both men. I think two things can be true, but the pub test of it is that you believe one or the other. And so I think that's where the public debate has gone and the public line of questioning has gone is, who do 13:15 you believe, Usman and his mates or Pat and his mates? 13:19 And the reality is there's probably an in-between there that is true, but no one really wants to admit to the extent to which that is true. 13:26 I'm Daniel James. Thanks so much for listening to 7am. 13:32 You
Community discussion