Who owns ideas in the AI age?

Can you ever really own an idea?

Recommended Video

The publishers, music producers, and film directors who make up the creative economy would say yes — as would many of the artists and writers they work with. But some in Big Tech are beginning to push back, arguing that ideas—like information—should be free, accessible, and repurposeable for anyone. When it comes to ideas, they argue, even those which spring directly from our own heads are the product of every other idea, environment, and person we’ve come into contact with. As such, they are fair game for training the large language models (LLMs) behind the AI platforms many of us have become reliant upon.

The argument has become increasingly urgent as generative AI companies build powerful models—and attract huge investment—by ingesting vast amounts of online text, images, and video, including books, journalism, and art created by humans.

This is the existential issue facing, among others, the international publishing giant Hachette. David Shelley, the company’s U.K. chief who also became U.S. CEO in January 2024, is joining the fight on behalf of creatives everywhere.

Shelley is a publisher through and through. The son of antique booksellers, he grew up above a bookshop and got his first industry role fresh out of university. You would be hard-pressed to find someone more passionate about, and invested in, the future of publishing. “We’re at an absolutely pivotal moment,” he says. “We need to stand up for the rights of the authors we work with and for the whole of the creative industries.”

Hachette vs. Google

This is not mere lip service. This January, Hachette asked a U.S. federal court for permission to intervene in a proposed class action lawsuit against Google. Along with Cengage, an education technology provider, the publisher claims the tech giant copied content from Hachette books and Cengage textbooks to train its large language model, Gemini, without asking permission. Google argues that training LLMs on vast text-based datasets is a transformative process which analyzes patterns in language, rather than reproducing the original works and, as such, qualifies as fair use.

Shelley isn’t buying it. “It’s just another form of theft,” he says. “We know these LLMs basically stole our authors’ work.”

This isn’t the first time Hachette has taken legal action against those looking to steal from it. In 2023, the company took on Internet Archive, an online library which offers users a free, digitized archive of music, books, and other publications. Hachette, along with Penguin Random House, HarperCollins, and Wiley, claimed the platform allowed people to download copyrighted books for free, against the authors’ wishes. In March 2026, Hachette Book Group, the American arm of the business, took on what it alleges is a pirate site, Anna’s Archive, for the same reasons.

Hachette has an impressive portfolio to protect. As one of the Big Five major global publishing houses, it is the force behind bestsellers from Donna Tartt’s The Goldfinch to Stephenie Meyer’s Twilight saga, as well as nonfiction titles such as Malcolm Gladwell’s Outliers and Mitch Albom’s Tuesdays With Morrie. Parent company Hachette Livre’s 2025 revenues exceeded €3 billion ($3.44 billion), driven by the work of popular authors across the 13 regions it operates in.

The Google lawsuit is just one of many examples of creatives taking on Big Tech. Across the U.S. and Europe, dozens of lawsuits have now been filed by individuals and organizations seeking to stop AI companies from training their models on copyrighted material without permission.

> 62%

Revenue growth since Shelley took the helm

> €3 billion

Total revenue for Hachette Livre in 2025

> 14%

Hachette’s share of the U.K. publishing market

Last year, three authors won a landmark victory against AI company Anthropic, resulting in a $1.5 billion settlement. It is worth noting, however, that they did not win on the grounds of breach of copyright. The judge ruled that Anthropic’s use of the authors’ work was “exceedingly transformative” and therefore allowed under U.S. law. Unfortunately for Anthropic, over 7 million of the books it had used to build its training library were pirated copies, each of which carried a potentially steep penalty.

For Shelley, this is really an issue of semantics. “Copyright and piracy often go hand in hand,” he says. He cites children’s writer Enid Blyton’s estate, which the publisher owns, as an example. “Blyton spent her whole life writing those books — that was her achievement. If you can then ingest those into an LLM and the model can use that to create copies, to me, it’s very clear that it’s her intellectual property that has been ingested and is being monetized.”

And here is the crux of the issue. Someone is making money from the use of these ideas—but it’s not the author, it’s the LLM companies. The commercial stakes are enormous: the global generative AI market was valued at $103.58 bi