OpenAI Fights Back Against Court Order Forcing It to Retain User Chats Indefinitely
OpenAI is challenging a federal court order that compels the company to preserve all user-generated content from ChatGPT and its API—including conversations users have deleted—arguing that the mandate clashes with its user privacy commitments.
The dispute stems from a legal battle with The New York Times, which sued OpenAI and Microsoft in December 2023.
The newspaper claims its copyrighted articles were used without permission to train the companies’ large language models, potentially violating intellectual property laws and threatening the financial sustainability of journalism.
Privacy Versus Legal Demands
On 13 May, a U.S. court ordered OpenAI to “preserve and segregate all output log data that would otherwise be deleted on a going forward basis.”
This applies to standard ChatGPT users across Free, Plus, Pro and Team tiers, as well as API users—excluding those on “Zero Data Retention” plans.
Enterprise and education accounts are not subject to the ruling.
OpenAI has objected strongly to the decision, warning it could undermine user trust.
Brad Lightcap, OpenAI's Chief Operating Officer, said in a statement on 3 June, when the company filed a request to vacate the order,
“We give you tools to control your data… This fundamentally conflicts with the privacy commitments we have made to our users.”
CEO Sam Altman also took to X, formerly Twitter, to express concern.
“We will fight any demand that compromises our users’ privacy; this is a core principle.”
Altman further questioned the broader implications:
“We have been thinking recently about the need for something like ‘AI privilege’; this really accelerates the need to have the conversation… talking to an AI should be like talking to a lawyer or a doctor.”
What Is The New York Times Looking For?
The Times has argued that if OpenAI continues deleting data, potential evidence of copyright infringement might be lost.
The paper’s legal team is seeking access to user interactions that may reflect outputs derived from its articles—possibly even in near-verbatim form.
These interactions could help support its claim that the AI model replicates and distributes its work without proper licensing or attribution.
In an earlier opinion from April, U.S. District Judge Sidney Stein acknowledged that The Times had provided “numerous” and “widely publicised” examples of ChatGPT responses closely mirroring its content.
He ruled that such examples were sufficient to allow parts of the lawsuit to proceed.
OpenAI Warns of Long-Term Impact
OpenAI maintains that indefinitely retaining user chats—particularly those unrelated to the lawsuit—sets a harmful precedent.
The company currently deletes conversations within 30 days after a user removes them, a policy it says aligns with global data protection norms.
According to OpenAI, the order jeopardises this practice and subjects millions of unrelated users to unnecessary data retention.
The company insists that all retained data is held securely, accessible only to a limited legal compliance team, and will not be shared with The Times unless required by further court orders.
If the appeal succeeds, OpenAI has pledged to return to its original data deletion practices.
Lightcap said,
“This is an overreach by The New York Times. We’re continuing to appeal this order so we can keep putting your trust and privacy first.”
A Broader Debate Over Fair Use
At the centre of the lawsuit is the legal doctrine of fair use.
The Times argues that the training of AI systems on its content—without a licence—goes beyond what is legally allowed.
It also claims that tools like ChatGPT and Bing Chat can summarise articles in ways that allow users to bypass its paywall, undermining its subscription model.
OpenAI has pushed back, accusing the newspaper of misrepresenting the facts.
The company said The Times had “cherry-picked” examples for its legal filings.
Altman said,
“We think this was an inappropriate request that sets a bad precedent.”
The order currently remains in place without a defined end date, as both sides continue their legal fight over copyright and user privacy.
Weiliang
Miyuki
Alex
Joy
Brian