The New York Times claims Microsoft encouraged OpenAI to use its articles

In short: The New York Times told a federal court that Microsoft actively supported OpenAI using Times articles, without permission, to train AI systems.

What happened

The New York Times is in an ongoing copyright lawsuit against OpenAI and Microsoft, filed on December 27, 2023 in federal court in New York. In new court materials, the Times argues that Microsoft did more than invest in OpenAI. It says Microsoft encouraged and benefited from OpenAI using Times journalism to train AI models.

Training an AI model means feeding it huge amounts of text so it can learn patterns in language, like a student reading a very large library to learn how to write. The Times says this “library” included millions of its articles, even though the company did not give permission or get paid.

The Times also claims the AI tools can sometimes produce output that closely matches Times writing, including content that normally sits behind a paywall. It says the tools have repeated parts of investigative stories and product advice from Wirecutter. The lawsuit also points to cases where the AI made up fake articles and credited them to The Times, which the publisher says can hurt its reputation.

Microsoft and OpenAI deny they did anything illegal. They argue the training is allowed under “fair use,” a part of U.S. copyright law that can permit limited use of copyrighted material in some situations. OpenAI has also argued that the Times had to use many unusual prompts to get some of the closest matches shown in court.

Why it matters

This case could help decide whether AI companies can learn from copyrighted news without licensing it, and what responsibility a partner like Microsoft has when it builds that AI into products people use every day.

Source: NYTimes