Our Take
Courts are starting to draw lines around AI company data requests, but this ruling turns on procedural grounds, not a sweeping privacy principle.
Why it matters
As litigation involving AI companies multiplies, the boundaries of what third parties can demand are still being written. This decision will be cited in future discovery disputes, making it worth tracking for anyone building or deploying AI in regulated industries.
Do this week
Legal counsel: document your data retention and user privacy policies now, before litigation lands, so you can defend discovery objections on grounds of proportionality and privacy, not surprise.
A federal judge declined to enforce a subpoena
In a lending-related lawsuit, a judge refused to compel OpenAI to hand over ChatGPT records requested by the plaintiff, according to Reuters. The specific factual basis for the refusal is not detailed in available reporting, but the decision hinges on whether the burden of producing the records outweighs their relevance to the case at hand.
This is a discovery ruling, not a final judgment on the merits. It addresses whether OpenAI must comply with a third-party subpoena during litigation, not whether OpenAI itself is liable or wrongdoing occurred.
Discovery fights are where legal exposure gets real
As AI companies face increasing litigation across employment, intellectual property, privacy, and consumer harm theories, they will be asked to produce internal records, training data, user interactions, and system logs. Courts are now signaling that blanket requests for AI platform records may not pass the proportionality test.
The ruling is narrow: it applies to this specific case. But it establishes precedent that judges can and will push back on expensive, broad discovery demands. For companies deploying AI, this means courts are unlikely to treat you as a neutral repository of all data you touch. The flip side: if your data practices are poor, a future judge may rule against you.
Legal and product teams should audit data retention now
If you operate an AI platform or service, document what user data you retain, how long you keep it, and why. Establish clear policies for data deletion, pseudonymization, and access controls. When a subpoena lands, you will want to show the court that you follow a principled retention schedule, not ad hoc storage.
Second: work with your lawyer to draft proportionality objections before they are needed. If you can articulate why producing all ChatGPT records (or your equivalent) is burdensome, costly, or contains privileged material, you are more likely to succeed in limiting scope. This ruling suggests judges are listening to those arguments.