OpenAI: The New York Times hacked ChatGPT to make it infringe copyright – IT Pro – News

According to OpenAI, The New York Times exploited a flaw in its AI models to apparently infringe the newspaper's copyright. The New York Times sued OpenAI in December, alleging that the company misused Times articles to train its chatbots.

It's called OpenAI In his official response to the New York Times' accusation that the newspaper paid someone to “hack” artificial intelligence models. This may be possible through exploiting a flaw and using “misleading claims,” though the company did not provide further details on this. Only this would make it possible to have ChatGPT answer multiple paragraphs of The Times articles word for word. Even then, it took “tens of thousands” of attempts to get the desired result, and the newspaper itself also fed parts of those articles to the chatbot, according to OpenAI.

The AI ​​company claims that users typically cannot use OpenAI simply to have Times articles presented to them at will. The paper is said to have shown “significant efforts” to ensure that the models “regurgitate” their training data directly. OpenAI claims that these are not full articles, but just parts of articles that can be read for free on third-party sites. Therefore, the company demands that four of the seven charges brought by the New York Times be dropped.

The New York Times' chief lawyer, Ian Crosby, is late To the record I know it's a “weird misconception” that the newspaper hacked the chatbot. According to Crosby, the New York Times “used OpenAI products only to prove that OpenAI stole and reproduced The Times’s copyrighted work.” Additionally, he called it “surprising” that OpenAI would admit that it monitors user claims, when the company said it does not. Finally, he finds it “remarkable” that the AI ​​company does not deny that it copied the newspaper's articles without permission.

In December, The New York Times filed a lawsuit against OpenAI and Microsoft for copyright infringement. The newspaper said at the time that the companies “illegally copied the Times's unique content” to train the AI. For some answers, ChatGPT “transcribes texts almost verbatim from NYT, which you would normally need a paid NYT subscription to obtain.” Meanwhile, by spreading false information, or “hallucinations,” chatbots could damage the New York Times’ image if they claim the information comes from the newspaper.