Federal Courts Divide on Production of Expert Witness AI Prompts
Courts are splitting over whether AI prompts used by experts to shape reports must be produced in discovery, creating new disclosure, work-product and sanctions exposure for litigants and experts.
Federal courts are reportedly splitting on whether prompt histories underlying an expert witness's work must be produced in discovery. On one side of the divide, courts have treated AI-generated work as no different from any other work product: if an expert used a tool in forming opinions, the inputs matter to testing methodology, bias, and reliability, and non-production risks a Rule 37(e) sanctions analysis. Other courts have leaned the other way, reasoning that prompts are not ordinary discoverable communications but privileged or immaterial scratch material, or that production threatens third-party confidentiality and trade secrets held by model providers.
The practical significance for major-firm litigators is immediate: expert engagement letters and disclosure obligations built around pre-AI assumptions may be under-specified, and prompt-log requests are now a live motion topic. Experts face deposition exposure about their tooling, and reliance vendors could be pulled in on confidentiality grounds.
Attention should stay on consolidation, appellate review, and any Rule 26 report rules addressing generative AI. Note that the underlying material here is a single firm alert; the specific decisions, courts, and dates are not established and should be verified against primary opinions before citing.