Microsoft says Copilot rarely copies NYT articles in copyright dispute

Microsoft says in new court filings that Copilot rarely reproduces even full sentences from copyrighted articles and books used to train its AI. In discovery, the company provided 8.2 million Copilot chats, finding 59,545 with at least 16 words in common to training content, while experts for CIR and the Authors Guild identified only a handful of substantial overlaps; The New York Times has disputed the conclusions as the copyright case against Microsoft and OpenAI moves forward.