More than 550 publishers have now joined a coalition suing OpenAI and Microsoft over the use of copyrighted material in training generative AI models. On Wednesday, 26 additional publishers became parties to the case, including Times Publishing Company, parent of the Tampa Bay Times; Austin Chronicle Corp.; and others.
The publishers allege that OpenAI and Microsoft used their work without consent or compensation to develop commercial AI products and removed copyright information including author names and copyright notices. Law firm Platkin LLP represents 60 of the publishers in the litigation.
Fair use at the center of the dispute
The core legal question is whether using copyrighted material to train large language models qualifies as fair use. Publishers argue that copying for model development cannot be classified as fair use and harms existing and potential markets. OpenAI and Microsoft contend that training is transformative because models learn to recognize patterns rather than reproducing original work.
On September 4, both sides filed competing requests for summary judgment in the federal case in the Southern District of New York. U.S. District Judge Sidney H. Stein invited amicus briefs by October 16. The U.S. Justice Department weighed in on September 2, supporting a broad interpretation of fair use in AI training while warning that compulsory licensing could disadvantage smaller companies.
The U.S. Copyright Office stated that fair-use determinations in AI training are fact-specific, depending on what material was used, how it was obtained, why it was copied, and its effect on copyright markets.
Mixed legal landscape
The legal picture remains uncertain. On September 16, OpenAI and Microsoft won a partial victory in the GitHub Copilot developer case, with an appeals court rejecting one copyright-management-information theory while leaving other claims intact.
The Copyright Office has indicated a preference for allowing voluntary licensing markets to develop before Congress considers compulsory licensing, while noting that licensing material at AI-training scale could create substantial financial and logistical burdens.
Market implications
Gartner forecasts that worldwide spending on AI models and platforms will rise from $39.31 billion in 2025 to $64.25 billion in 2026, a 63.4% increase. Foundation generative AI models alone are projected at $23.36 billion.
Any shift toward paying for training data access could substantially increase costs in this rapidly growing market. The OECD warns that concentrated access to data, computing power, and cloud infrastructure can strengthen larger incumbents, and copyrighted training data could become another costly input that larger developers are better positioned to absorb.
The legal outcome will affect not only how AI companies source training data but potentially determine whether licensing costs widen or narrow the competitive field for building next-generation models. Similar issues are emerging globally, with the European Union developing frameworks for AI copyright obligations under its AI Act.
