Publishing grapples with where to draw the line on AI

Useful as a tool for research, generative AI is also posing a threat to authors with a model that's 'good at creative writing' in the works.

- Andrew Hill

Since columnist Parmy Olson won 2024's Financial Times and Schroders Business Book of the Year Award with Supremacy, about tech companies' battle for control of artificial intelligence (AI), she has started using large language models (LLMs) more frequently in her own research.

"(They) can be a helpful tool for bouncing ideas around, (exploring) angles and getting historic references to make comparisons," she says.

As the 2025 edition of the prize launches, the debate over whether generative AI is a threat or an opportunity for authors is consuming the industry.

"We're keenly aware these technologies can be used in ways that will dilute the market for human-authored works," says Mr Umair Kazi, director of advocacy and policy at The Authors Guild, the US professional organisation for writers. "But at the same time, they are hugely useful tools."

Generative AI's fluent prose might put some writers out of a job. Evidence is also growing that some LLMs have been trained by developers—without authors' consent—on pirated versions of copyrighted books.

Concern about illegal scraping has united authors against the practice. Ms Mary Rasenberger, a former copyright and media lawyer who is now chief executive of The Authors Guild, says "we have never before had that level of agreement among our membership on any issue".

The challenge of AI has also brought together publishers and agents.

Mr Esmond Harmsworth, president of the literary agency Aevitas, says: "Since the author and the publisher could easily be replaced (by AI), it's been a more pleasant negotiation and one in which we join forces to try to come up with solutions to this."

Agents are now insisting on clauses in book contracts to control the future training of LLMs on authors' work or, in some cases, license its use for a fee.