Prompts, privilege & discovery
In re OpenAI, Inc. Copyright Infringement Litigation (ChatExplorer)
The court denied a request for all prompts and outputs from The New York Times’s internal infringement-testing tool as irrelevant and disproportionate.
Discovery denied01
Case background
Material facts
OpenAI and Microsoft sought the Times’s ChatExplorer prompts and outputs, plus testimony about several missing months of test data, to support fair-use and non-infringing-use theories.
02
The decision
The court’s ruling
The magistrate judge found the requested testing material insufficiently relevant and disproportionate. Because it was not discoverable, its loss did not trigger Rule 37(e).
Why it matters
What the decision means
Not every prompt history becomes discoverable merely because AI behavior is central to the lawsuit. The requesting party still must connect the specific tests to a claim or defense and justify the burden.
Limit of the ruling. The defendants objected. No later written decision reversing this order was located by the September 5, 2026 cutoff.
Primary reading