Microsoft says fewer than 1 percent of 8.2 million Copilot chats shared 16 words with the news those chats pulled in
Microsoft provided 8.2 million Copilot chat logs in a copyright fight with news publishers and book authors, Lauren Feiner at The Verge writes. Microsoft says fewer than 1 percent of those chats shared at least 16 words with news content used as source material for the answer.
Discovery is the dig through each other’s files before trial. Microsoft says it gave publishers the logs most likely to contain their work, chosen because they hit keywords implicating use of the news plaintiffs’ websites.
The resulting analysis, per Microsoft via The Verge, shows 59,545 of the 8.2 million chats contained at least 16 words in common with that news source material. An expert for the Center for Investigative Reporting found 51 instances of substantial overlap with CIR work in the same dataset, Microsoft says.
In the authors’ case, an expert found only 24 responses with at least 30 matching words across the 8.2 million conversations. Only 10 of 212 books evaluated had any matches, Microsoft claims.
Microsoft argues the numbers support fair use for training. Fair use is a legal defense that says some uses of copyrighted work without a license are allowed when the new product serves a different purpose. Microsoft says occasional reproduction hardly undermines the transformative purpose of training large language models.
The New York Times disagrees. Lead counsel Ian Crosby said documents and testimony from discovery lead to one conclusion: Microsoft and OpenAI stole from the Times to make commercial products that substitute for its journalism. CIR and the Authors Guild did not immediately comment to The Verge.
Microsoft filed Friday seeking summary judgment, an early ruling that would end the case without a full trial. Publisher and author claims were consolidated under one judge despite objections. These are filings and company framing, not a finding.
Eight million chats. Under one percent at sixteen words. The question is whether that is the tell Microsoft wants, or the tell the Times already rejects.