Microsoft has reportedly reached an agreement with HarperCollins, a subsidiary of News Corp., to utilise select nonfiction titles for training its artificial intelligence models.
In a statement to Bloomberg News, HarperCollins confirmed it reached an agreement that would “allow limited use of select nonfiction backlist titles for training AI models to improve model quality and performance”.
The deal includes strict limitations and safeguards to protect author rights and royalty streams.
The agreement, yet to be publicly detailed, enables Microsoft to access HarperCollins' nonfiction backlist titles to improve AI model quality and performance.
HarperCollins noted that authors could opt out of the initiative, ensuring their rights are respected.
"Part of our role is to present authors with opportunities while protecting the underlying value of their works and shared revenue streams," HarperCollins said in a statement. The publisher emphasised that the deal includes "clear guardrails" around the use of AI outputs.
Technology firms, including Microsoft, are increasingly seeking high-quality text for AI training to enhance model accuracy and expertise.
Publishers like HarperCollins are becoming pivotal partners in this pursuit.
News Corp., HarperCollins’ parent company, previously signed an agreement with OpenAI, allowing content from outlets like The Wall Street Journal and Barron’s to be used for AI training.
Microsoft has also partnered with publishers such as Reuters, Axel Springer, and Hearst for similar initiatives.