OpenAI is challenging a court directive requiring the indefinite retention of all ChatGPT output data, arguing that it compromises users’ privacy and contradicts its data handling policies.
The case results from a copyright lawsuit filed by The New York Times, which accuses OpenAI and Microsoft of using millions of its articles to train ChatGPT without authorisation.
The court, in response, ordered OpenAI to preserve and isolate all user outputs generated by ChatGPT, potentially including deleted or expired conversations.
On 3 June, OpenAI formally asked U.S. District Judge Sidney Stein to reverse the preservation order issued in May, describing the demand as excessive and damaging to user trust. OpenAI maintains that complying with such a tough data retention requirement clashes with industry norms and existing user agreements.
“This fundamentally conflicts with the privacy commitments we have made to our users. It abandons long-standing privacy norms and weakens privacy protections,” said Brad Lightcap, chief operating officer at OpenAI.
OpenAI currently allows users, particularly those using ChatGPT Free, Plus, Pro, and Team accounts, to delete chats, which are then scheduled for permanent removal from its servers within 30 days.
The same rules apply to API users unless they are operating under Zero Data Retention agreements, where no data is logged or retained at all.
The Times, however, has pushed for indefinite data storage, contending that user interactions with ChatGPT could potentially expose copyrighted material that was improperly reproduced. OpenAI rejects this approach, calling it speculative and harmful.
“We will fight any demand that compromises our users’ privacy; this is a core principle,” said OpenAI CEO Sam Altman in a post on X. “We think this was an inappropriate request that sets a bad precedent.”
The court’s preservation order also includes deleted data, which OpenAI says it would normally purge under its standard policy. To comply, the company has had to establish a separate, secure legal hold system, accessible only to a limited and audited legal and security team.
Importantly, OpenAI clarified that data stored under this legal hold is not automatically shared with The New York Times or any other party. Any future attempt to access the data would be met with legal resistance.
Users on ChatGPT Enterprise, ChatGPT Edu, and API customers using Zero Data Retention endpoints are unaffected by the order.
So far, the court has only partially narrowed the scope of the data order. During a hearing in late May, it was confirmed that ChatGPT Enterprise data is exempt. OpenAI continues to appeal, hoping the broader directive will be overturned.
The lawsuit, filed in 2023, accuses OpenAI and Microsoft of copyright infringement. Judge Stein, in an earlier opinion, stated that The New York Times had shown credible evidence that the two companies could be liable for encouraging users to replicate protected content through the chatbot.
He noted multiple examples of ChatGPT reproducing portions of Times articles without attribution or licensing.
OpenAI, while not denying that outputs occasionally resemble public content, insists that these occurrences are exceptions, not deliberate reproductions. They argue that complying with the court’s order would force them to betray the very users whose trust they rely on to maintain their platform.
There is currently no end date or review period for the preservation order. For now, OpenAI is bound by the court’s mandate but says it is doing everything possible to limit the long-term impact on its users.