'David and Golliath' moment as writers union protest at Meta HQ over 'theft' of work

In March 2025, reports alleged that over 7.5 million books were pirated to train Meta's new AI system. Now writers are fighting back for compensation and to protect the future of the arts

Writers union Society of Authors have protested outside Meta HQ in London on April 3 in response to the news that thousands of books have been ‘stolen’ by the media giant to train their AI system, Llama 3. In March 2025, The Atlantic released a searchable database of millions of books and research papers that have allegedly been used by Meta to train AI.

Society of Authors CEO Anna Ganley spoke to The Mirror ahead of the protest. She said that this was a "David and Golliath" moment as writers and creatives react to their work being stolen by big tech. The scale of the issue is vast, as The Atlantic’s data set, called Library Genesis or ‘LibGen’, contained over 7.5million books and 81 million research papers that were affected.

In the UK, fair dealing laws protect creatives from unlawful usage of copyrighted work. Ganley said: “Authors’ works have been stolen. And big tech is getting away with it.” The writers union is demanding a meeting with both the UK Government and with Meta to seek compensation for UK authors.

Meta Platforms, Inc is the parent company of social media companies Facebook, Instagram and WhatsApp, and holds a market value £1.147 trillion. This is in stark contrast to the decline in author income in the United Kingdom.

Ganley tells The Mirror: “Median income for authors from books sales is £7,000 per year, it is worse that authors’ works are pirated and available illegally - for free - online in shadow libraries that lurk on the web – but it is downright criminal that big tech has scraped the web – and our authors’ works - for their own commercial gain.”

Under the UK fair dealing exception to copyright, there is no question that scraping is unlawful without permission. Ganley explained that if this usage of writers works goes unchecked that it will have a long term effect on the arts industry. She said: “This unlicensed use of creative works poses an existential threat to the livelihoods of creators and authors everywhere.”

A UK Government spokesperson for the Department for Culture, Media and Sport said in response: “The status quo is holding back the creative industries, media and AI sector from realising their full potential — and that cannot continue.”

But this goes beyond UK borders, with the breach occurring in the United States. The UK Government official told us that following a consultation they will be adopting a “new approach” to licensing content for use, which they hope will enable greater transparency and lawful access to material to train world-leading AI models in the UK.

They continued: “As things stand, creatives are not being fairly compensated for their content and only the largest rights holders have the power to effectively assert their rights through licensing deals and the courts. AI developers are choosing to train their models in other countries with different copyright rules.”

Nobel Prize winning writer Sir Kazuo Ishiguro is one of the writers affected. “If someone wants to take a book I’ve written and turn it into a TV series, or to print a chapter of it in an anthology, the law clearly states they must first get my permission and pay me,” he explains. “To do otherwise is theft.”

Richard Osman, the best-selling author of novels such as The Thursday Murder Club, has had his works ‘scraped’ in the AI training system, too. He said: “Copyright law is not complicated at all. If you want to use an author's work, you need to ask for permission. If you use it without permission, you're breaking the law. It's so simple.” On the battle ahead for fair usage, he explains: “It'll be incredibly difficult for us, and for other affected industries, to take on Meta, but we'll have a good go.”

The Mirror reached out to Meta for their response. A spokesperson for Meta said: “We respect third-party intellectual property rights and believe our use of information to train AI models is consistent with existing law.”