NYT says OpenAI hid evidence in ChatGPT copyright case
The New York Times and The Daily News claim OpenAI failed to turn over important evidence in their lawsuit. The dispute centers on whether ChatGPT was trained on their journalism and whether its answers repeated that journalism. OpenAI has argued that searching its and customer would be hard and would raise privacy concerns. A court-ordered deposition reportedly showed that OpenAI privacy engineer Vinnie Monaco said the company had already run internal searches and checks on its for ed news content.
The s also claim OpenAI had built a database of about 78 million ChatGPT conversations before the lawsuit and later used a inside a tool set called Project Giraffe to detect when answers repeated protected material. The s first asked for a sample of 120 million , but the number was reduced to 20 million. OpenAI submitted that smaller sample last December, but it reportedly had so many s that the court saw it as unusable. The s also claim OpenAI deleted billions of ChatGPT outputs after the lawsuit began, despite a court order to preserve them, and replaced millions of logs in the requested sample.
They want the judge to punish OpenAI by blocking the 20 million log sample from evidence, treating major repetition of their content as established, limiting OpenAI’s arguments about the sample, and making OpenAI pay related legal costs. OpenAI denies the claims and says the Times is trying to reach private conversations from people unrelated to the case while OpenAI defends user privacy and fair use.
Key points
- The s say OpenAI hid or mishandled evidence in the ChatGPT lawsuit.
- OpenAI had said searching and would be difficult and raise privacy issues.
- The s claim OpenAI already had about 78 million conversations and tools to detect repeated material.
- A 20 million chat log sample was reportedly so heavily redacted that the court treated it as unusable.
- OpenAI denies the allegations and says it is defending user privacy and fair use.