Court Orders Production of Expert AI Prompts

Magistrate Judge Thomas O. Farrish ordered the plaintiff to produce the artificial intelligence prompts used by expert Dr. Naomi Oreskes in Conservation Law Foundation v. Shell Oil Co., as reported by Reason and Arnold Porter. The court found the prompts are part of the expert's methodology and therefore discoverable under Rule 26, and it declined to interpret a Rule 29 discovery agreement as broadly shielding the prompts from disclosure, per Arnold Porter and Law360. The opinion quoted by Reason emphasized that a Rule 29 agreement "must be quite clear" before it will supersede otherwise-relevant discovery. The court also addressed the plaintiff's assertion that only "search terms" were used and that no additional responsive materials existed, according to Arnold Porter.

Industry reporting and practitioner guidance, including an Arnold & Porter synthesis, group recent case law into distinct categories: attorney-crafted prompts used in furtherance of litigation, prompts generated by parties or clients without counsel direction, prompts used by expert witnesses as part of methodological work, and AI use on discovery materials governed by protective orders (Arnold Porter; Lex-Arca). Editorial analysis: Observers note that these categories produce different privilege and discovery outcomes, with attorney work product often receiving stronger protection when tool architecture and terms of service do not expose data to third parties.

Editorial analysis: For litigators and practitioners working with AI, the ruling reinforces that prompt-level actions can be treated like any other methodological step in expert workups and therefore may be subject to disclosure. Reporting in Law360 and CCB Journal frames this decision as one of the first to squarely address expert prompts, adding to the emerging case law that governs when prompts are discoverable versus protected. Editorial analysis: Tool architecture, vendor terms of service, and whether prompts are created by counsel or an expert will increasingly determine whether prompts are privileged, a point emphasized in practitioner writeups cited by Lex-Arca and Arnold Porter.

Editorial analysis: Observers will track follow-on district court opinions and any appellate treatment that clarifies how courts draw lines between protected attorney work product and discoverable expert methodology for AI-assisted analysis. Editorial analysis: Parties and their counsel are likely to refine preservation, logging, and protective-order strategies, and courts will pay attention to evidence-based challenges to a party's claim that no responsive materials exist, as the Farrish opinion relied on that principle when ordering further disclosure (Arnold Porter; Reason).

Editorial analysis: When AI is used to filter, prioritize, or analyze document productions, practitioners should expect that prompts and prompt-derived artifacts can be treated as part of an expert methodology and therefore potentially discoverable in litigation; tool choice and documentation practices affect the legal exposure, according to the emerging practitioner literature and firm guidance cited above (Lex-Arca; Arnold Porter).