Based on reports from Bloomberg, Microsoft has reached an agreement with HarperCollins, a subsidiary of News Corp., to use select non-fiction titles from the publisher's catalog for training its artificial intelligence models, marking a significant development in the intersection of publishing and AI technology.
The collaboration between HarperCollins and Microsoft represents a strategic move in the evolving landscape of AI and publishing. Under this agreement, Microsoft will gain access to a selection of non-fiction titles from HarperCollins' extensive catalog to train its upcoming AI models12. This partnership underscores the growing importance of high-quality, curated content in developing sophisticated AI systems. While the specific details of Microsoft's AI project remain undisclosed, the deal highlights the tech giant's commitment to enhancing its AI capabilities using reputable literary sources34.
HarperCollins has become the first of the "Big Five" U.S. trade publishers to enter into an AI licensing agreement, marking a significant milestone in the publishing industry's engagement with artificial intelligence1. This groundbreaking deal allows for the use of select nonfiction backlist titles for AI model training, with a focus on improving model quality and performance23. The agreement's key features include:
A three-year licensing period for the selected titles1
An opt-in system for authors, respecting their rights to choose participation4
Clear guardrails to protect authors' rights and shared revenue streams5
Limited scope of use, specifically for AI training purposes4
This move by HarperCollins sets a precedent for other major publishers and highlights the growing intersection between traditional publishing and emerging AI technologies15. The deal not only represents a new revenue stream for publishers and authors but also raises important questions about the future of content licensing and intellectual property rights in the age of AI67.
HarperCollins has implemented an opt-in system for authors, offering $2,500 per title for a three-year licensing period12. This arrangement includes protections regarding credit and limits on verbatim usage, ensuring authors maintain control over their work's involvement in AI training. The deal specifically focuses on select non-fiction titles, with clear boundaries and guardrails in place to safeguard authors' rights, income, and royalties34. By allowing authors to choose whether to participate, HarperCollins aims to balance innovation in AI development with the protection of creators' interests in the rapidly evolving digital landscape.
The HarperCollins-Microsoft agreement reflects a broader trend in the tech industry, where major companies are seeking partnerships to secure high-quality training data for their AI models. This deal follows similar recent agreements, such as OpenAI's partnerships with various media organizations, valued at $250 million, and Google's $60 million agreement with Reddit for AI training access12. These collaborations highlight the increasing importance of diverse and authoritative content sources in developing more sophisticated and capable AI systems, as tech giants compete to enhance their artificial intelligence offerings.
The agreement between HarperCollins and Microsoft has sparked diverse reactions within the writing community. Notably, author and comedian Daniel Kibblesmith publicly criticized the deal, describing it as "abominable" and urging readers to support authors by purchasing physical books from local bookstores1. This response highlights the ongoing debate surrounding AI's role in the literary world and the potential impact on authors' livelihoods. Despite the controversy, HarperCollins maintains that the agreement's limited scope and clear guardrails effectively protect authors' rights and interests2.