OpenAI ‘accidentally’ deletes evidence of plagiarism in New York Times articles

OpenAI allegedly accidentally deleted evidence related to a copyright infringement case. Data found on website New York Times And Used as evidence in million dollar legal dispute.

In December 2023, New York Times He accused OpenAI of “copying and using” millions of articles published on the tool to train large language models (LLM). In this process, Teams Answers from ChatGPT and Copilot (both powered by models from the GPT family) were highlighted, where entire quotes were copied from publications or the literary style was imitated.

On Wednesday (20) OpenAI reported: evidence was accidentally deleted. According to the company’s lawyers, the dismissal was due to a “glitch”.

“On November 14, 2024, plaintiffs were informed that OpenAI engineers had deleted all programs and data created by the investigation and stored in private virtual machines,” the tool’s defense said.

defense NYT I don’t believe this is a mistake

defense NYT he “sees no reason to believe” the hypothesis that this was a mistake.

OpenAI tried to recover the data but All saved content is incomplete and not 100% reliable. “On November 19, plaintiffs realized that the recovered data did not have the original structure or names and could not be used to determine whether vehicle parts were used to develop defendant’s models,” the plaintiff’s defense said. The statement was included.

To move forward the legal dispute NYT he would have already invested US$1 million (R$5.8 million in direct conversion). The outlet expects OpenAI to be held liable for “billions of dollars in legal damages” for the alleged plagiarism.