News · 2026-08-01

A judge did not rule that ChatGPT users have no rights to their chats

A federal magistrate in New York denied one ChatGPT user permission to join the copyright litigation against OpenAI, and the order says the data preservation hold in that case exists for a possible spoliation inquiry rather than to hand user conversations to the New York Times. Viral summaries claiming a court declared that users have no rights in their own chats misstate what was decided. Separately, a Minnesota law restricting AI-generated intimate imagery took effect on 1 August.

Key facts

• Magistrate Judge Ona T. Wang denied Aidan Hunt's motion to intervene in the consolidated OpenAI copyright litigation, citing the absence of a direct legally protectable interest and procedural defects.
• The order states the preservation requirement was tied to a possible spoliation inquiry, not to producing conversations to the New York Times.
• Minnesota's provider-facing prohibition allows penalties up to $500,000 per unlawful access, download or use.
• Primary sources: the filed order, Doc. 688 and Minnesota Session Laws 2026, Chapter 72.

There is a genuine privacy story here. It is just not the one that went around.

What happened: in the consolidated case captioned The New York Times Company v. Microsoft Corporation et al, an individual user asked the court to let him participate, arguing his interests were at stake because his conversations fell within a data preservation order. Motions to intervene succeed when the person has a direct, legally protectable interest in the specific action. A copyright dispute between publishers and an AI company is about whether training and outputs infringe; it is not adjudicating what rights users hold in their chat histories. The court said so, and also flagged procedural problems with how the motion was brought.

The difference between "a user cannot join this copyright case" and "users have no rights in their conversations" is the difference between being told you cannot testify at someone else's trial and being told you have no legal standing anywhere. The first is routine case management. The second would be extraordinary. Only the first happened.

The order also corrects a second misunderstanding. A preservation order tells a party to stop deleting data so that evidence is not destroyed while litigation is live - the legal equivalent of taping over the delete button on a security recorder during an investigation. It does not, by itself, transfer anything to the opposing party. The court's own language ties the requirement to a potential spoliation inquiry, which is the question of whether relevant material was destroyed, not a disclosure route to the Times.

The real caveat is in OpenAI's retention policy rather than in the docket, and it is worth reading carefully. OpenAI says deleting a chat removes it from your account immediately and schedules permanent deletion from its systems within 30 days - except where the data has already been de-identified and disassociated from your account, or must be retained for security or legal reasons. Those two exceptions do a lot of work. Clicking Delete does not retroactively de-identify a conversation that has already been through that pipeline, and it does not release data held under a legal obligation. OpenAI has separately said the litigation hold ended for new data and for conversations originating in Europe, Switzerland and the UK, while some historical data from April to September 2025 remains under secure legal hold.

So the accurate summary is uncomfortable in a different way than the viral one. No court has stripped users of rights. But the practical reach of your delete button is bounded by de-identification pipelines and legal holds you cannot see, and the person best placed to explain those bounds is the company holding the data.

The Minnesota development is a separate and more concrete change. As of today, the state prohibits a person who owns or controls a website, application, software or other service from letting users realistically generate or alter an identifiable person to depict intimate parts, with a carve-out for work requiring substantial individualised technical or artistic skill. It is enforceable through private suits and attorney-general penalties reaching $500,000 per unlawful access, download or use. Federal court intake records list a case captioned X.AI LLC v. Ellison, No. 0:26-cv-03425, assigned to Senior Judge Donovan W. Frank - a challenge to the statute. The underlying motion papers and any ruling were not available for review, so nothing should be assumed about how a judge has treated it.

The honest caveat: this is a report on documents, not on outcomes. The intervention denial is procedural and could be revisited on a proper record; the Minnesota challenge is live; and the retention exceptions described above are OpenAI's own account of its systems, not an audited finding. What is settled is narrow - a court declined to let one person into a case, and said its data hold was not a disclosure order - and that narrow thing is the opposite of what circulated.

Primary source, verified: read the paper →

Ground Truth. (2026, August 1). A judge did not rule that ChatGPT users have no rights to their chats. Ground Truth. https://groundtruth.day/news/a-judge-did-not-rule-that-chatgpt-users-have-no-rights-to-their-chats.html

@misc{groundtruth:a-judge-did-not-rule-that-chatgpt-users-have-no-rights-to-their-chats,
  title  = {A judge did not rule that ChatGPT users have no rights to their chats},
  author = {{Ground Truth}},
  year   = {2026},
  month  = {aug},
  url    = {https://groundtruth.day/news/a-judge-did-not-rule-that-chatgpt-users-have-no-rights-to-their-chats.html}
}

Topics: cybersecurity · privacy · data-retention · legal · openai · policy