Source: Microsoft signs a deal with News Corp's HarperCollins to use nonfiction titles to train an unannounced AI model; HarperCollins says authors can opt out
- Nonfiction books will be used for training AI models — AI companies and publishers have butted heads in lawsuits
Context & Ripple Effects
The agreement marks a commercial route through a conflict that had already prompted News Corp and other publishers to join a coalition pressing for AI rules and litigation. It also sits alongside unresolved copyright and fair-use questions in cases involving Microsoft and OpenAI, covered in legal analysis of the New York Times case.
First-order effects
- Microsoft gains a permissioned pool of HarperCollins nonfiction material for its unannounced model, while HarperCollins turns part of its catalog into a licensable AI asset.
- Eligible HarperCollins authors get an opt-out choice, making participation—not simply publisher ownership—a practical constraint on the training corpus.
Second-order effects
- The deal gives other publishers and AI developers a concrete licensing-and-author-consent model to negotiate against, rather than relying solely on disputed training-data claims.
- Compensation and contract terms become more consequential for authors and agents; later coverage of the debate over payments for book licensing shows that access to titles does not settle how proceeds are divided.
Third-order effects
- If such deals proliferate, high-quality published text may increasingly be controlled through negotiated rights, shifting AI training-data access toward publishers with catalog scale and rights-management infrastructure.
- Opt-out provisions could make author consent a recurring feature of book-data licensing, though their influence will depend on uptake and on how copyright disputes develop.
The trend: AI developers are moving from contested access to copyrighted text toward negotiated, rights-managed content supply chains.