TL;DR: On May 18, 2026 a District of Connecticut magistrate judge ordered the plaintiff in Conservation Law Foundation v. Shell Oil Co. to produce artificial intelligence prompts used by its expert to generate a key report. The ruling treats AI prompts as part of the expert’s methodology under Rule 26(b)(3) and signals a surge in prompt-level discovery practices that trial teams must anticipate, manage, and defend against with robust protective orders and disciplined AI workflows. (bowmanandbrooke.com)
What happened in Shell Oil Co. v. Conservation Law Foundation
Conservation Law Foundation, Inc. v. Shell Oil Co., No. 3:21-cv-00933 (D. Conn.), reached a milestone in the discovery of AI-assisted expert work. On May 18, 2026, Magistrate Judge Thomas O Farrish granted a motion to compel production of reliance materials, including AI prompts and queries that the plaintiff’s expert used to develop her opinions. The court rejected the argument that AI prompts lay outside the scope of a Rule 29 discovery agreement and treated the expert’s AI-driven methodology as a discoverable component of the expert’s work product. In practical terms, the order requires CLF to disclose the prompts used by its expert Naomi Oreskes, framing the prompts not as “notes” tucked away in a separate file, but as an integral part of the expert’s methodology under Rule 26(b)(3). This marks one of the clearest statements to date that prompt-level artifacts can be fair game in civil litigation where AI tools assist expert analysis. (bowmanandbrooke.com)
The Connecticut docket and contemporaneous reporting confirm that the ruling is not merely rhetorical; it directs production of AI prompts and cautions about how Rule 29 discovery agreements interact with AI workflows. Several external analyses describe the decision as a practical blueprint for handling AI inputs in expert work, including protective-order tweaks to limit disclosures and to require retention conditions aligned with training-data concerns. While Shell Oil is a single, high-profile environmental matter, the decision squarely addresses a question that recurs across civil and regulatory litigation: when does AI-assisted analysis cross into discoverable material? (leagle.com)
For context, the Shell Oil ruling sits in a broader wave of early 2026 decisions addressing AI, privilege, and discovery. A May 28, 2026 roundup from a major firm notes parallel developments in SDNY, E.D. Mich, and D. Colo, including how courts treat AI-generated materials under the attorney-client privilege and work-product protection, and how protective orders are being crafted to address AI data retention and training concerns. Those discussions underscore the practical trajectory: AI prompts, outputs, and related workflows are becoming routine subjects of discovery and protective-order negotiations, not esoteric edge cases. (kirkland.com)
Why this matters to trial teams
Prompt-level discovery is here. The Shell Oil order makes clear that the inputs that an expert used to generate a report—down to the prompts themselves—can be discoverable as part of the expert’s methodology. In civil litigation especially, this means counsel should be prepared to defend the confidentiality and strategic architecture of AI-assisted work, not merely the outputs. (bowmanandbrooke.com)
Protecting the work product is still a living concept with limits. The decision reinforces that while AI prompts may be discoverable as methodology, the court did not automatically compel the disclosure of the AI tool’s identity. This creates a divide that practitioners must navigate: preserve identity where disclosure would reveal strategic thinking, but disclose prompts that reveal the approach used to narrow or analyze data. Expect more protective-order language aimed at this balance. (kirkland.com)
Protective-orders must evolve with technology. The Shell Oil case, together with other contemporaneous rulings, shows courts educating parties on how to tailor protective orders to AI workflows, including prohibiting input of confidential information into open AI platforms unless strong data-privacy safeguards are in place, or requiring “closed” enterprise systems with clear data-retention controls. Those guardrails are likely to become standard in discovery practice. (kirkland.com)
The practical impact for trial readiness. For trial teams, these developments elevate the importance of documenting, defending, and testing AI workflows well before trial. The ability to defend the use of AI in preparing expert reports hinges on pretrial planning about what can and cannot be disclosed, how prompts are archived, and how disclosure triggers affect privilege and trial strategy. (kirkland.com)
Practical steps for trial teams in light of this ruling
Revisit protective orders now. Ensure protective orders explicitly address AI prompts, including language that restricts input of Confidential Information into consumer-grade AI tools, demands zero-data-retention clauses where appropriate, and requires claw-back or deletion provisions at the conclusion of the matter. The Morgan v. V2X decision from Colorado in March 2026 provides a useful blueprint for cross-jurisdiction thinking about AI in protective orders, including the balance between protecting materials and acknowledging legitimate discovery requests. (kirkland.com)
Create auditable AI workflows. Start an auditable trail of AI inputs, including prompts and document-assistance steps used by experts. While the identity of the tool may be protected in some contexts, producing a transparent methodology log can reduce disputes later about what the expert did, why, and how AI guided conclusions. This aligns with the Shell Oil approach of treating prompts as part of the expert’s methodology under Rule 26(b)(3). (bowmanandbrooke.com)
Engage early with the opposing side on AI discovery scope. Expect early discovery disputes focused on AI prompts, inputs, and training data, and be ready to negotiate scope with concrete examples (for instance, particular prompt sets linked to a specific analysis). Firms tracking these cases suggest that courts are receptive to carefully tailored protective orders and to limits on the dissemination of AI-input materials. (kirkland.com)
Plan cross-examination around AI prompts. Anticipate how prompts and AI-driven analytics will be used at trial and prepare objections or foundations that address the reliability and relevance of AI-assisted conclusions. The Shell Oil decision demonstrates that prompt-level materials can influence expert credibility and the admissibility of AI-derived opinions.
Objection Academy: training for the AI-era trial lawyer
The rise of AI prompt discovery makes robust objections and evidence-application training more essential than ever. Objection Academy offers a suite of tools aimed at sharpening trial advocacy through objection drills, evidence-law training, and courtroom simulations. For practitioners, these features translate into practical preparation for AI-related evidentiary issues at trial:
See Objection Academy as a core practice aid for objection-driven trial-readiness, including drills that cover the admissibility and foundation of AI-assisted evidence and expert methodology. The platform emphasizes real-time courtroom pressure and feedback to improve objection timing and precision. (objectionacademy.com)
The platform’s Trial Simulator provides AI-generated decision points that mimic a live trial narrative, enabling attorneys to rehearse how to handle AI-driven lines of questioning and the introduction of AI-supporting materials in direct and cross. This can be especially valuable as courts increasingly scrutinize AI prompts and outputs in expert reports. (objectionacademy.com)
The app and related content also highlight FRE and evidence training that remains relevant as the legal landscape evolves with AI. A dedicated guide on the best app for Federal Rules of Evidence in 2026 positions Objection Academy as a strong option for evidence training in this new era. (objectionacademy.com)
California and New York attorneys may find MCLE credits available through Objection Academy’s offerings, a practical consideration for maintaining ongoing compliance while building trial-readiness skills. (objection.com)
By integrating such training into pretrial preparations, trial teams can translate the Shell Oil order’s practical implications into a disciplined, repeatable process for handling AI-induced discovery and evidence issues in future matters.
What to watch next
Courts continue to refine the boundaries around AI in discovery and privilege. The Shell Oil decision sits alongside other 2026 decisions that are shaping protective-order language and discovery scope for AI workflows. Attorneys should monitor subsequent rulings from the District of Connecticut and other districts on AI prompts, tool identity, and the weight of AI-generated materials in expert reports. The evolving framework is likely to drive more explicit, machine-generated evidence gatekeeping and to encourage standard templates for AI-related discovery and protective orders. (kirkland.com)
Sources
Conservation Law Foundation, Inc. v. Shell Oil Co., Civil No. 3:21-cv-00933 (D. Conn.). Order granting motion to compel production of reliance materials, May 18, 2026. Justia docket and public records, District of Connecticut. (law.justia.com)
Shell Oil Co. discovery prompts order discussion and analysis. Bowman & Brooke LLP, May 2026. (bowmanandbrooke.com)
A Federal Court Charts a Path on AI, Protective Orders and Work Product in Discovery. Kirkland & Ellis, May 20, 2026. (kirkland.com)
Federal Courts Issue Diverging Rulings on the Use of Generative AI in the Context of Privilege, Work Product and Protective Orders. Akin Gump, May 28, 2026. (akingump.com)
Objection Academy: Best app and features for trial attorneys in 2026. Objection Academy site, including Trial Simulator and MCLE information. (objectionacademy.com)