Microsoft says Copilot rarely reproduces NYT texts — The Verge
Microsoft said that its Copilot chatbot only rarely reproduces entire sentences or substantial fragments of materials from The New York Times, other media outlets and books. The company cited these data in court filings in a copyright infringement case brought against Microsoft and OpenAI by publishers and authors, The Verge reports.
Analysis of 8.2 million conversations
As part of discovery, Microsoft provided an expert retained by news publishers with 8.2 million Copilot conversation logs. According to the company, the records were specifically selected using keywords related to the use of the plaintiff news outlets' websites, so this sample was the most likely to contain their works.
Microsoft said that 59,545 records contained at least 16 words shared with news content on which the AI model was based. This amounts to less than 1% of all 8.2 million records. An expert from the Center for Investigative Reporting, Microsoft claims, identified 51 instances of a “substantial match” with the organization’s materials in the sample.
More current news is available on the UA.News Telegram channel Telegram.
The company’s position in the dispute
In the lawsuit concerning books, an expert retained in the authors’ case found 24 Copilot responses containing at least 30 words matching the texts, Microsoft reports. The company also said that matches were found for only 10 of the 212 books assessed.
Microsoft believes these results support its argument that the use of copyrighted materials to train artificial intelligence systems constitutes fair use. According to the company, Copilot and similar systems use such content for a substantially different purpose than the original publications. The occasional reproduction of excerpts, according to Microsoft’s position, does not negate the transformative nature of training large language models.
Publishers and authors argue that Microsoft and OpenAI created products based on their works and now compete directly with them, including by reproducing protected content. The lawsuits by news publishers and book authors were assigned to one judge to streamline the proceedings. Microsoft is asking the court for summary judgment, which could end the case at an early stage. The New York Times, the Center for Investigative Reporting and the Authors Guild did not immediately comment on the company’s statements.
