Microsoft: Our Copilot chatbot rarely copies news articles
Microsoft claims its Copilot AI assistant almost never reproduces full sentences or substantial chunks from news articles. The company presented this evidence in a legal battle with The New York Times and other publishers.
Microsoft is fighting a copyright lawsuit from The New York Times, other news organizations, and book authors. They say these publishers claim Copilot — Microsoft's AI-powered chatbot — is copying their work without permission. In court documents, Microsoft argues this simply isn't happening.
To prove it, Microsoft handed over 8.2 million chat logs (records of conversations users had with Copilot) to an independent expert hired by the news publishers. A chat log is basically a written record of what someone typed to the AI and what it replied. Microsoft specifically selected logs that were most likely to contain copied material — conversations that included keywords from news websites.
The analysis found something surprising: out of nearly 60,000 conversations examined, Microsoft says Copilot almost never produced even complete sentences from the original articles, let alone entire passages that would serve as a substitute for reading the real article. In other words, when people used Copilot, they still needed the original news sources to get the full story.
This dispute matters because it touches on a bigger question: as AI systems become more powerful, how much can they learn from published content without owing money to the creators? The outcome could shape how news organizations, authors, and AI companies work together in the future.
Original source: The Verge AI
