Image: Adobe Stock
When 800 million people use ChatGPT weekly for their most personal digital conversations, a fight like this does not stay in the courtroom.
Grab that popcorn. The privacy battle between OpenAI and The New York Times is a remarkable one.
OpenAI has released a statement accusing the newspaper of wanting to invade the privacy of its users as their copyright lawsuit intensifies into something far more concerning for everyday users.
When 800 million people use ChatGPT weekly for their most personal digital conversations, a fight like this does not stay in the courtroom. It lands in everyone’s lap.
What began as a training data copyright dispute is now a privacy brawl. According to OpenAI’s announcement, The Times initially asked for 1.4 billion ChatGPT conversations. OpenAI whittled that to 20 million random conversations, December 2022 through November 2024.
Pause on that. Twenty million private chats, any of which could hold raw, personal details from users around the world.
The newspaper’s lawyers pointed to another AI company that had already turned over five million private chats in an unrelated case. The subtext was not subtle: if they did it, OpenAI should too.
OpenAI’s Chief Information Security Officer Dane Stuckey fired back, writing that “Trust, security, and privacy guide every product and decision we make”, and the company is asking the court to reject The Times’ demand entirely.
The Times says it might find examples of users trying to get around their paywall by asking ChatGPT to reproduce articles. To find those needles, you have to sift the haystack, which means peering into how people actually talk to AI.
The lawsuit alleges that OpenAI gave Times content particular emphasis when building their LLMs, revealing a preference that recognizes the value of those works.
A Times spokesperson told media outlets that “No ChatGPT user’s privacy is at risk” because a court ordered anonymized samples under a protective order. Still, anonymized data can often be re-identified with shocking ease, and that reassurance rings hollow to many.
This controversy flows from a May 13, 2025 preservation order that upended OpenAI’s data retention policies six months ago. Deleted ChatGPT conversations were once permanently erased within 30 days. Now they are preserved indefinitely.
The order covers ChatGPT Free, Plus, Pro, and Team subscriptions, and it excludes ChatGPT Enterprise. In practice, millions of individual users lost privacy protections while business customers kept theirs.
When the decision hit social media, it sparked immediate concern among users who had trusted ChatGPT with sensitive details, from finances to relationship problems to health worries. Late-night confessions, meet legal hold.
OpenAI says it is minimizing the damage with a “de-identifying procedure” that scrubs personal information from conversations. The preserved data is stored separately in a secured system, accessible only to a small, audited legal security team.
Even so, the promise changed. Users who deleted chats expecting them to vanish now face indefinite preservation, with the possibility of future legal scrutiny.
The ripple effects are already here. Judge Stein affirmed the preservation order during a June 26, 2025 hearing four months ago, signaling that AI chats may not be truly private when legal disputes erupt.
Companies are rethinking AI data governance. Employees who used personal ChatGPT for work questions might have unknowingly put employer information under a legal hold. Organizations are scrambling to map their exposure and adopt Zero Data Retention agreements that prevent prompts and responses from being stored at all.
The case exposes a bigger mismatch. Traditional privacy frameworks were not built for an AI era where a single order can pull hundreds of millions of conversations into potential discovery.
OpenAI is still fighting the data demands, and The Times is still pushing for broader access. Expect more filings and regular updates from OpenAI as the standoff continues.
For now, conversations from April through September 2025 sit in legal limbo, securely stored but not turned over to any party. OpenAI says that if it succeeds in challenging the order, it will resume standard deletion practices.
But trust is fragile. This is the first time a court mandated mass preservation of AI-generated content at this scale, and it changes how people think about AI privacy. The genie is out of the bottle.
Research by Wiz shows that AI firms, with combined valuations exceeding $400 billion, have left the equivalent of their front doors propped open.
