Microsoft argues Copilot rarely reproduces copyrighted news and books in copyright dispute

Microsoft is defending itself against copyright infringement lawsuits by arguing that Copilot seldom reproduces meaningful passages from articles or books, positioning the company's use of protected content as a transformative practice that falls within fair use doctrine.

The Numbers Behind Microsoft's Defense#

As part of ongoing discovery in the case brought by The New York Times, book authors, and the Center for Investigative Reporting, Microsoft shared 8.2 million Copilot chat logs with experts hired by the publishers. The company selected these logs specifically because they contained keywords most likely to trigger matches with the plaintiffs' original work.

According to Microsoft's analysis of the dataset:

• 59,545 conversations (roughly 0.7%) contained at least 16 consecutive words matching news content
• 51 instances of substantial overlap with Center for Investigative Reporting material were identified
• 24 responses in the authors' evaluation contained at least 30 matching words across 8.2 million interactions
• Only 10 of 212 books showed any word-match overlap

The Fair Use Argument#

Microsoft contends that these figures demonstrate the company's use of copyrighted material during AI training qualifies as fair use. The company argues that although large language models require training on protected content, the resulting systems serve fundamentally different purposes than the originals. In its court filing, Microsoft states that occasional text reproduction "hardly undermines the transformative purpose of LLM training."

Next Steps in the Legal Battle#

The case—consolidated under a single judge to handle both the publishers' and authors' claims despite their initial objections—will hinge on whether courts accept Microsoft's fair use logic. Microsoft submitted the filing Friday while requesting summary judgment, which would resolve the case without going to trial. The plaintiffs claim that Microsoft and OpenAI built competing products directly on their works and now reproduce protected content in real time. The Trump administration has also filed a statement supporting OpenAI in the case this week.