OpenAI has discontinued its practice of indefinitely preserving most ChatGPT users’ deleted and temporary chat logs, marking the end of a contentious preservation order that emerged from ongoing copyright litigation with major news organizations.
The change follows months of legal wrangling that placed user privacy concerns at the center of a high-stakes copyright dispute between artificial intelligence companies and traditional media outlets.
Court Order Required ChatGPT Log Retention Despite Privacy Concerns
The preservation mandate originated from a lawsuit filed by news organizations including The New York Times, alleging that ChatGPT users were circumventing paywalls by requesting AI-generated summaries of copyrighted articles. Plaintiffs argued that users attempting to avoid detection would likely configure their conversations as temporary or delete chat histories afterward.
OpenAI initially contested the preservation order, maintaining that the requirement conflicted with its privacy policies and user expectations regarding data retention. Despite these objections, the court ruled against the company, compelling OpenAI to retain ChatGPT output logs beginning earlier this year.
By July, news organizations had begun examining the preserved logs, which contained only ChatGPT’s responses rather than complete conversation histories. Multiple attempts by ChatGPT users to formally intervene in the case were rejected, as courts determined they lacked standing as non-parties to the litigation.
Partial Data Preservation Continues Under Modified Agreement
US Magistrate Judge Ona Wang approved a joint motion from news plaintiffs and OpenAI on Thursday to terminate the broad preservation order, though monitoring provisions remain in place for specific user accounts.
Under the revised agreement, OpenAI ceased preserving “all output log data that would otherwise be deleted” as of September 26. However, the arrangement includes significant carve-outs that maintain access to previously collected information.
All deleted and temporary chats saved during the preservation period remain accessible to news organizations, who are searching for instances of copyright infringement or misattribution of false information to their publications. This archived data provides plaintiffs with a substantial repository for identifying potential violations.
Additionally, OpenAI continues monitoring ChatGPT accounts associated with domains flagged by news organizations during their data review process. If plaintiffs identify additional suspicious domains in future coordination meetings with OpenAI, those accounts may be added to the ongoing monitoring program.
Copyright Litigation Expands Beyond User Data Disputes
The conflict between OpenAI and news organizations extends well beyond user log preservation, with recent developments indicating escalating legal complexity.
Co-defendant Microsoft has filed motions to exclude its AI companion Copilot from the litigation, attempting to separate its products from OpenAI’s legal exposure. The maneuver highlights how copyright disputes involving AI training data are creating ripple effects across the technology sector.
The litigation carries substantial implications for both parties. News organizations contend that ChatGPT and similar AI tools pose existential threats to their business models by potentially replacing them as information sources. Beyond revenue concerns, publishers worry about reputational damage resulting from AI systems attributing fabricated information to their outlets.
OpenAI faces mounting pressure that may ultimately force settlement negotiations, though not necessarily from legal opponents. Insurance providers have reportedly declined to offer comprehensive coverage for AI products facing multiple potentially multibillion-dollar lawsuits, creating financial incentives for OpenAI to resolve pending litigation.
The preservation order dispute illustrates the tension between AI innovation and established intellectual property frameworks. As generative AI systems become increasingly capable of producing content similar to copyrighted material, courts and companies continue wrestling with questions about appropriate data handling, user privacy, and copyright protection.
The modified preservation agreement represents a compromise that partially addresses privacy concerns while maintaining investigative access for copyright holders. However, the underlying copyright questions driving the litigation remain unresolved, ensuring continued legal battles over how AI companies train and deploy large language models.