OpenAI said Monday that a New York Times NYT-N lawsuit against it was “without merit” and that it supported and created opportunities for news organizations, as it waded further into a debate over the unauthorized use of published work to train artificial intelligence technologies.
The Times sued OpenAI and Microsoft MSFT-Q on Dec. 27, accusing the companies of infringing on its copyrights by using millions of its articles to train AI technologies such as the ChatGPT chatbot. Chatbots now compete with the Times as a source of reliable information, the lawsuit said.
In a 1,000-word blog post Monday, OpenAI said it collaborated with news organizations and had struck partnerships with some of them, including the Associated Press. Using copyrighted works to train its technologies is fair use under the law, the company added. The Times’s lawsuit does not tell the full story of how OpenAI and its technologies operate, it said.
“We look forward to continued collaboration with news organizations, helping elevate their ability to produce quality journalism by realizing the transformative potential of A.I.,” the company wrote.
Lindsey Held, a spokesperson for OpenAI, declined further comment.
The Times was the first major American media organization to sue OpenAI and Microsoft over copyright issues related to its written works. Other groups, including novelists and computer programmers, have also filed copyright suits against AI companies. The suits have been spurred by the boom in “generative AI,” technologies that generate text, images and other media from short prompts.
OpenAI and other AI companies build this technology by feeding it enormous amounts of digital data, some of which is likely copyrighted. That has led to a realization that online information – stories, artwork, news articles, message board posts and photos – may have significant untapped value.
AI companies have long claimed that they can legally use such content to train their technologies without paying for it because the material is public and they are not reproducing the material in its entirety.
In its blog post, OpenAI said its discussions with the Times about a potential partnership appeared to progress constructively, with a last communication Dec. 19. During the negotiations, it said, the Times had mentioned that it had seen OpenAI’s technology “regurgitate” some of its content – meaning the technology had generated near-verbatim excerpts from articles that ran in the Times – but declined to provide examples. When the Times sued eight days later, OpenAI said it was surprised and disappointed.
The Times didn’t immediately respond to a request for comment.
OpenAI said its technology sometimes regurgitates articles, but that was a “rare bug” that it was working to solve. The Times’s lawsuit included examples showing ChatGPT reproducing excerpts from its articles nearly word for word.
“Intentionally manipulating our models to regurgitate is not an appropriate use of our technology and is against our terms of use,” OpenAI said.