Microsoft and HarperCollins strike licensing deal to train AI models on nonfiction books

Microsoft has signed a licensing deal with HarperCollins to train its AI models.

According to Bloombergsources, HarperCollins will allow Microsoft's LLMs to train on nonfiction titles. Microsoft reportedly doesn't plan on creating AI-generated books, but instead will access high quality data to make its models more intelligent and accurate. "HarperCollins authors will have the option to participate or not," said the outlet.

SEE ALSO: ChatGPT now saves chat history even if you've opted out of sharing training data

404 Media first broke the news of a licensing deal with an anonymous AI company. Author Daniel Kibblesmith shared screenshots of an email on Bluesky, likely from his agent informing him about the deal. "You are likely aware, as we all are, that there are controversies surrounding the use of copyrighted materials in the training of AI models," said the memo. "Much of the controversy comes from the fact that many companies seem to be doing so without acknowledging or compensating the original creators. And of course there is concern that these AI models might one day make us all obsolete."

Mashable Light Speed Want more out-of-this world tech, space and science stories? Sign up for Mashable's weekly Light Speed newsletter. By signing up you agree to our Terms of Use and Privacy Policy. Thanks for signing up!

Abominable.

[image or embed]

— Daniel Kibblesmith (@kibblesmith.com) November 15, 2024 at 2:36 PM

According to the screenshots of the email, HarperCollins is offering a non-negotiable payment of $2,500 per title for a three-year licensing deal.

HarperCollins confirmed there is a deal with an unnamed AI company, telling Bloomberg, "its limited scope and clear guardrails around model output" respects authors' rights while presenting them with new opportunities.

Meanwhile, multiple outlets have reported that AI companies, like Google, OpenAI, and Anthropic are getting diminishing returns with the development of new models because they're running out of high quality data to train on. Microsoft was not included in these reports, but its Copilot model relies on underlying genAI technology from OpenAI. So training AI models on nonfiction works might be a strategy to combat lessening improvements.