According to Mark Price, there are “two grounds” on which a legal case could realistically be pursued against Meta in the UK for the company’s alleged use of pirated books to train artificial intelligence (AI) models. Just two you might say, that seems too few for such an egregious act. Lord Price, a former MD of Waitrose, has written to Meta CEO Mark Zuckerberg asking why “it is acceptable to ignore property rights and take my work, and that of thousands of other British authors, for Meta’s commercial benefit”, and has called on other writers to join his action.

Price’s lawyers contend that the first way would be to understand whether books taken from the pirate database LibGen have been “ingested and processed within the UK”, which Price claimed would determine if there is a UK case against Meta. But it is the second way that might provide for some interesting toing and froing: Price wants to see whether the content that is generated by Meta’s Llama differs significantly from the works used to train these models. “What we may be able to claim is that, if you ask Llama for information, it regurgitates something that’s incredibly close to what you had in your book,” Price told The Bookseller.

In that regard, he is watching closely the Getty Images case against Stability AI over that company’s alleged use of copyrighted images, which will go to trial in June 2025. The claimants have said that the content produced by the company’s AI model, Stable Diffusion, reproduced parts of the images that were used during the training to a significant extent.

Meta has argued in the US that there is no proof of economic damage to authors

Meta has argued in the US that there is no proof of economic damage to authors because their work is never reproduced by its AI model – just used for training – but if Getty Images can prove otherwise then it opens up a whole new legal front. When HarperCollins signed its licensing arrangement with an AI business – thought to be Microsoft – it made restrictions on how much of a book could be seen as part of the arrangement. But, with no such agreement in place with rightsholders, Meta should be wary about what its AI puts out. Already, a former Meta lawyer has admitted that this perhaps unintended consequence of the machines taking over could be what does for the firm in copyright courts.

I have written about this subject before, and recently, and make no apologies for that.

In the UK this week, the Data (Use and Access) Bill will be used as another moment to firm up copyright law. Amendments to be debated during the third reading in the House of Commons will attempt to ensure that there is compliance, transparency and enforcement, which – if approved – should additionally curb the UK government’s attempts to grant tech companies exceptions around their use of published materials to train AI – a position the government should have come to anyway.

According to Tom West, CEO of the Publishers’ Licensing Services, writing on The Bookseller’s website this week, the bill could “turbo charge” the licensing of content. “The call for proper accountability is not anti-tech nor anti-innovation, rather it is a recognition that, as generative AI plays an increasingly significant role in our lives, the accuracy and quality of information has never been more important,” West argues.

And he is right. We are at an inflection point. If losing control results in harm, chaos or actions you later regret, it is usually a sign that boundaries or regulation are required. Not my words, of course, but those of ChatGPT.