Particle.news
Download on the App Store

OpenAI Moves to Overturn Order for 20 Million De‑Identified ChatGPT Logs as Deadline Nears

The fight turns on whether de‑identified chat logs under a protective order sufficiently safeguard users’ privacy to allow testing of the news outlets’ copyright claims.

Overview

  • Magistrate Judge Ona T. Wang’s Nov. 7 order requires OpenAI to produce a random sample of consumer conversations by Nov. 14, 2025, or within seven days of completing de‑identification.
  • OpenAI filed to vacate the ruling, warning that producing entire conversations risks exposing sensitive details and asserting that more than 99.99% of the logs are irrelevant to the case.
  • The sample spans December 2022 through November 2024 and, according to OpenAI, excludes enterprise and other business customers, with access restricted to vetted experts under a protective order.
  • The New York Times and other news plaintiffs argue the logs are necessary to assess whether ChatGPT reproduced their content and to rebut OpenAI’s claim that they manipulated outputs.
  • The court relied on de‑identification safeguards and cited the Concord v. Anthropic discovery approach, against a backdrop of earlier preservation orders that extended OpenAI’s data retention beyond its 30‑day policy.