The New York Times Sues OpenAI and Microsoft Over Using Its Articles to Train AI

The newspaper says millions of its articles were copied without permission to build ChatGPT, which now competes with it for readers, and it wants billions in damages and the destruction of any program trained on its work.

The New York Times filed a lawsuit in federal court in Manhattan today against OpenAI and Microsoft, accusing them of copyright infringement on a massive scale. The complaint says the companies used millions of Times articles, without permission or payment, to train the programs behind ChatGPT and the new Bing, and that those programs now compete with the newspaper by producing its reporting for free. It asks for damages it does not specify but describes as "billions," and for the destruction of any program built on the Times's work.

The complaint's most striking pages show ChatGPT reproducing Times articles nearly word for word when asked — long passages of a 2019 investigation of the taxi industry, for instance — and describes cases where the chatbot attributed to the Times things it never published. The newspaper says it tried to negotiate a licensing agreement for months before suing.

OpenAI's position, stated in other cases, is that learning from public writing is "fair use" — the same doctrine that lets a person read a library and write a book — and that its programs do not store or reproduce articles but learn patterns from them. It has signed licensing deals with some publishers, including the Associated Press, and says it wants to with the Times.

Authors, artists, and comedians have sued the AI companies this year. This is the first major news organization to do so, and the one with the most resources to see it through. If the Times wins, every AI program trained on the open internet was built on stolen material, and the industry's economics change overnight. If it loses, the internet is a free training set. The court will take years. The question won't wait.