Microsoft Partners With HarperCollins For Nonfiction AI Training

The tech giant Microsoft and HarperCollins, one of the world’s largest English-language publishing companies, signed a deal to train artificial intelligence model with data from nonfiction titles.

According to Bloomberg, the book publisher will allow Microsoft’s software to learn information from its books to train an AI model—that has not been disclosed yet—but not to generate books. An anonymous source with knowledge of the deal shared the information with Bloomberg.

Microsoft’s role in this deal was revealed by Bloomberg yesterday, right after a heated debate between authors and HarperCollins and after 404 Media shared that a secret tech company was behind the AI training. Microsoft declined to comment on the new deal when Bloomberg reached out.

American writer and comedian Daniel Kibblesmith shared on Bluesky that HarperCollins offered him a fixed price of $2,500 to opt in and allow AI to use his creative content on his book Santa’s Husband—a fictional children’s book suggesting that the deal could go beyond non-fiction. The author was outraged and many users joined the debate.

It’s not the first time this year that a big tech company has partnered with a large publisher. In May, OpenAI signed a multiyear agreement with News Corp—HarperCollins’s parent company—to access multiple publications like the New York Post, The Wall Street Journal, and The Daily Telegraph with a focus on journalism.

In June, it was revealed that Microsoft has been training a new large language model (LLM), known as MAI-1, to compete with Google and OpenAI. The tech giant has been working on multiple AI developments and expanding its chatbot capabilities. In September, Microsoft announced multiple enhancements to its AI agent Copilot and new features like Copilot Pages and interactions with programs like PowerPoint, Excel, and Outlook.