Deleted ChatGPT conversations are now at the center of a high-stakes copyright fight. After OpenAI objected to a court order requiring it to retain all ChatGPT logs “indefinitely,” including deleted and temporary chats, US district judge Sidney Stein denied the company’s objections.
The result is a difficult position for OpenAI: it must preserve data it says users expected to be protected, while negotiating how news plaintiffs led by The New York Times may search a limited sample of those retained logs.
Why the ChatGPT logs are being preserved
The retention order was issued by magistrate judge Ona Wang only days after news organizations, led by The New York Times, asked for it. The plaintiffs argued that the order was needed quickly to preserve potential evidence in their copyright case.
Their concern is that ChatGPT users may delete chats in which they attempted to use the chatbot to skirt paywalls to access news content. The order therefore covers logs that otherwise might not have remained available, including deleted and temporary chats.
OpenAI argued that the order forced it to abandon “long-standing privacy norms” and weaken privacy protections that users expected under ChatGPT’s terms of service. Stein was not persuaded. He suggested that OpenAI’s user agreement allowed data to be retained as part of a legal process, and said that is what is happening now.
What the search process may look like
OpenAI has said through a spokesperson that it plans to “keep fighting” the order. But the company appears to have limited paths left. One possibility described in the source is asking the Second Circuit Court of Appeals for a rarely granted emergency order, though OpenAI’s spokesperson declined to confirm whether the company will pursue that option.
For now, the more immediate issue is how the retained data will be searched. OpenAI and the news plaintiffs are negotiating a process that would allow searches through the retained data while limiting the scope.
The source indicates that The New York Times and other news plaintiffs are unlikely to search everything. Instead, only a small sample will likely be accessed, based on keywords agreed to by OpenAI and the plaintiffs. The data is expected to remain on OpenAI’s servers, where it will be anonymized, and will likely never be directly produced to the plaintiffs.
Both sides appear to want to reduce the amount of time the logs remain preserved. That creates a practical tension: moving ahead with some data collection could help bring retention to an end sooner, while continuing to fight the order could extend the period in which more private conversations are preserved.
Why the logs matter in the copyright case
For OpenAI, access to the logs carries litigation risk. The source says the logs could reveal examples of infringing outputs that might increase damages in the case. They could also show how often outputs attribute misinformation to news plaintiffs.
For the news plaintiffs, the logs are not described as central to the case. They may provide additional examples of copying, but the more important argument may involve market effects. The plaintiffs could use the logs to argue that ChatGPT dilutes the market for their content, which could weigh against OpenAI’s fair use argument.
That matters because a judge in a recent ruling said evidence of market dilution could tip an AI copyright case in favor of plaintiffs. The source does not say that the logs will prove that point, only that they could help news organizations make the argument.
Privacy concerns extend beyond this case
Jay Edelson, described as a leading consumer privacy lawyer, told Ars that he is concerned judges are not giving enough weight to user privacy. He argued that any evidence in the ChatGPT logs would not “advance” the news plaintiffs’ case “at all,” while changing “a product that people are using on a daily basis.”
Edelson warned that OpenAI itself probably has better security than most firms, but said lawyers have a weaker reputation for securing data. He said the idea that lawyers could be handling “some of the most sensitive data on the planet” should make people uneasy.
The largest user risk, according to Edelson, is a data breach. But that is not the only concern. Corynne McSherry, legal director for the digital rights group the Electronic Frontier Foundation, previously told Ars that retained user data could also be exposed through future law enforcement and private litigation requests.
Edelson also criticized the way the order treats different users. From court filings, it appears possible that enterprise users were excluded to protect OpenAI’s competitiveness, but Edelson said there is “no logic” to that exclusion “at all.” He argued that excluding enterprise customers may have removed the users with the strongest resources to challenge the order.
In his view, the order affects ordinary users most directly. He said it is “only going to intrude on the privacy of the common people out there,” and criticized Wang’s decision to deny two ChatGPT users’ request to intervene.
What users can take from the dispute
The source makes clear that most users’ chats are unlikely to be included in the eventual sample. But the broader issue is that chats users believed could be deleted or kept temporary are now being preserved because of litigation.
That possibility may change how some people use AI tools. Edelson said the threat of inclusion alone could push users to rethink how they use AI, and suggested that users moving to services such as Anthropic’s Claude or Google’s Gemini could mean the order is influencing market forces.
The dispute is still moving through negotiation and possible further challenge. But one point is already clear from the order’s survival so far: private AI conversations can become relevant to litigation in ways users may not expect, even when the searches are limited, anonymized, and kept on company servers.