The New York Times (NYT) stated in court that OpenAI engineers accidentally deleted data that the newspaper's team had been extracting from the ChatGPT system for 150 hours. This data was planned to be used in court as potential evidence. This was reported by WIRED.
Although most of the data was recovered, the NYT legal team notes that the original file names and folder structure were lost. Because of this, it is impossible to determine where and how OpenAI could have used the copied articles.
The NYT filed a lawsuit against OpenAI and Microsoft in December 2023. The publication accused both companies of using their materials to train artificial intelligence. In its turn, the AI company called the lawsuit groundless. Currently, the case is still at the discovery stage, and all parties are submitting the documents required to collect evidence.
As part of this case, the developer of ChatGPT was forced to show the newspaper the data used to train large language models, which had previously been kept secret. To test this data, OpenAI created a sandbox of two virtual machines where NYT lawyers could search for data.
Specialists spent more than 150 hours collecting the necessary data before it was erased on one of the devices. Despite the attempts to restore it, the publication's lawyers note that specialists will have to start collecting data from the beginning, which will require a lot of human resources. At the same time, the lawyers also note that they have no reason to believe that the data was deleted on purpose and call this situation a "failure."