#CaseoftheWeek with Kelly Twigger: Conservation Law Foundation V. Shell Oil

Why It Matters

The decision establishes that AI‑generated prompts are discoverable, forcing parties to negotiate AI‑specific protections and shaping the future of both climate‑adaptation litigation and broader AI‑driven discovery practices.

Key Takeaways

• AI prompts used by expert are discoverable under Rule 26.
• Magistrate’s order is a minute docket entry, not a published opinion.
• Rule 29 stipulations must explicitly name AI artifacts to be protected.
• Shell’s climate‑adaptation suit hinges on expert’s AI‑filtered documents for case.
• Florida Supreme Court also amends AI accountability rules this week.

Summary

The Meet and Confer podcast examined the Conservation Law Foundation’s lawsuit against Shell Oil, focusing on a May 18, 2026 minute order that ruled AI prompts used by the plaintiff’s expert witness are discoverable under Rule 26. The order, issued by Magistrate Judge Thomas Farish, arrived amid a broader debate over generative‑AI use in litigation and was subsequently stayed pending a district‑court review.

The court treated the AI prompts as part of the expert’s methodology, citing prior discovery of technology‑assisted review protocols. It rejected CLF’s three arguments: that prompts are outside discovery, that a Rule 29 agreement covering “notes” shielded them, and that the protective order barred disclosure. The decision aligns with earlier rulings in Heppner, Gilbarco and Morgan, extending AI‑prompt discovery to testifying experts.

Judge Farish emphasized that “AI prompts are an aspect of the expert’s method,” echoing language from Machia v. ADP. The expert, Harvard historian Naomi Oreskes, employed OpenAI’s ChatGPT‑4.0/4.5 via a secure Microsoft Azure API to cull millions of Shell documents. The case also coincided with a Florida Supreme Court amendment to Rule 2 of its general practice rules, signaling parallel state‑level moves on AI accountability.

For litigators, the ruling underscores the need to explicitly name AI queries, outputs, embeddings, and model parameters in Rule 29 stipulations and protective orders. Failure to do so leaves AI artifacts vulnerable to discovery, potentially reshaping strategies in high‑stakes environmental suits and any matter where experts rely on generative‑AI tools.

Original Description

Kelly Twigger of Minerva26 discusses 𝐂𝐨𝐧𝐬𝐞𝐫𝐯𝐚𝐭𝐢𝐨𝐧 𝐋𝐚𝐰 𝐅𝐨𝐮𝐧𝐝𝐚𝐭𝐢𝐨𝐧 𝐯. 𝐒𝐡𝐞𝐥𝐥 𝐎𝐢𝐥 (𝐃. 𝐂𝐨𝐧𝐧. 𝐌𝐚𝐲 𝟏𝟖, 𝟐𝟎𝟐𝟔), where Magistrate Judge Thomas O. Farrish ordered plaintiff to produce the AI prompts and queries its expert witness used to cull defendants' document production, holding that an expert's AI methodology is fair game for Rule 26(b) discovery and rejecting the argument that prompts qualify as protected "notes" under the parties' Rule 29 agreement. #caselaw #ediscovery #CaseoftheWeek

Link to case: https://app.minerva26.com/case_law/70240-conservation-law-found-inc-v-shell-oil-co?q=&counsel=

Link to Minerva26 Blog: https://minerva26.com/blog/

Link to podcast: https://minerva26.com/podcasts/