Publishers Push Licensing

British publishers are trying to turn AI training from a scraping fight into a licensing business, and they say the market is expanding fast. (publishers.org.uk) The Publishers Association said on March 3 that publishers have licensed content for text and data mining for a decade, that AI training deals were in place by 2023, and that licensing for retrieval-augmented generation, or RAG, is now a growing part of the market. (publishers.org.uk) The same report said the number of publishers active in AI licensing is set to almost double in 2026 and is expected to include all major academic publishers by year-end. Cambridge University Press backed that view on March 4 and said many AI developers are already using licensed content in the UK. (publishers.org.uk) (cambridge.org) In publishing, a training licence lets an AI company pay for access to books, journals or articles to build a model, while a RAG licence covers systems that fetch and summarize content at query time instead of memorizing it during training. The Professional Publishers Association said on April 22 that most of the 63 deals it tracked were RAG agreements. (sr.ithaka.org) (ppa.co.uk) The push comes as Britain is still deciding how copyright law should apply to AI developers. The government’s March 18 report to Parliament said it was reviewing policy options after a consultation that ran from December 17, 2024, to February 25, 2025, including licensing, transparency and enforcement around AI training. (gov.uk) Publishers say they want the government to reject any broad copyright exception for AI training and force developers to come to the table. The Publishers Association’s February 2025 consultation response said AI models should be developed on a “legal, sustainable, ethical and secure” basis, and Cambridge said the industry wants clearer rules on unlicensed use. (publishers.org.uk) (cambridge.org) The commercial case is already visible in disclosed deals. Informa’s Taylor & Francis gave Microsoft nonexclusive access to academic content in a deal worth $10 million upfront plus recurring payments, while Wiley reported $29 million in AI licensing revenue for the year ended April 30, 2025, up from $23 million a year earlier. (insidehighered.com) (publicnow.com) News publishers have also signed direct agreements with AI firms. News Corp and OpenAI announced a multi-year partnership on May 22, 2024, giving OpenAI permission to display content from outlets including The Wall Street Journal and use it to enhance products. (newscorp.com) (openai.com) Not every publisher can negotiate one by one, so collective licensing is starting to emerge. Publishers’ Licensing Services launched an opt-in scheme on March 10 with the Copyright Licensing Agency and the Authors’ Licensing and Collecting Society to let AI companies buy access under standard terms and licence fees. (thebookseller.com) Publishers are also responding to a traffic problem, not just a copyright one. The Professional Publishers Association said AI bots now account for one visit for every 31 human visits to publisher sites, up 60% in a year, and publishers expect search traffic to fall by as much as 40% over the next three years. (ppa.co.uk) Authors and academics have raised objections to some deals, arguing they were not properly consulted or compensated. The Society of Authors said in December 2024 that AI training rights belong to authors, while scholars told Inside Higher Ed and Times Higher Education they were surprised by Taylor & Francis’s Microsoft agreement. (societyofauthors.org) (insidehighered.com) (timeshighereducation.com) For AI companies, the direction of travel in Britain is straightforward: more publishers are asking for contracts, more intermediaries are building licensing pipes, and the era of treating publisher archives as free fuel is getting harder to defend. (publishers.org.uk) (thebookseller.com)