---
title: "Your Expert’s AI Logs Are Now Discoverable"
description: "Key quote: An expert witness's methodology is fair ground for discovery, see, e.g., Macchia v. ADP, Inc., 711 F. Supp. 3d 162, 167-68 (E.D.N.Y. 2024), and under the facts of this case, the process by..."
url: https://kaynemcgladrey.com/your-experts-ai-logs-are-now-discoverable/
date: 2026-05-28
modified: 2026-05-28
author: "Kayne"
image: https://kaynemcgladrey.com/wp-content/uploads/2026/05/May-28-Motion-to-Compel_1.webp
categories: ["Articles"]
type: post
lang: en
---

# Your Expert’s AI Logs Are Now Discoverable

!(https://kaynemcgladrey.com/wp-content/uploads/2026/05/May-28-Motion-to-Compel_1-768x195.webp)

**Key quote**:

> An expert witness's methodology is fair ground for discovery, see, e.g., Macchia v. ADP, Inc., 711 F. Supp. 3d 162, 167-68 (E.D.N.Y. 2024), and under the facts of this case, the process by which Dr. Oreskes culled down the defendants' document production into a subset to be worked with is an aspect of that methodology.

**Why it matters**:

If you are relying on AI to sift through evidence during a breach investigation or litigation, you need to stop assuming those inputs are safe. On May 18, 2026, Magistrate Judge Thomas O. Farrish in (https://www.courtlistener.com/docket/60042265/conservation-law-foundation-inc-v-shell-oil-company/) (3:21-cv-00933) declared that prompts are fair game by accepting a (https://storage.courtlistener.com/recap/gov.uscourts.ctd.145184/gov.uscourts.ctd.145184.941.0_1.pdf) (PDF) filed late last year. The (https://www.courtlistener.com/docket/60042265/conservation-law-foundation-inc-v-shell-oil-company/?page=7#entry-970) the plaintiff to hand over the specific AI prompts used by expert Dr. Naomi Oreskes, ruling they are part of her discoverable methodology under Rule 26(b), and not protected drafting notes.

The fatal flaw wasn't the technology; it was the paperwork. CLF tried to argue the expert only used "search terms," but the court saw right through it. Dr. Alexander Kaurov, the expert's own research assistant, had already referenced "prompt" in a declaration. That single slip-up gave the defendants an evidence-backed reason to doubt CLF's claim that nothing more existed. As Judge Farrish noted, a good-faith assertion that materials don't exist fails when the record contradicts it.

This changes the risk calculations for CISOs and counsel. We've known that Slack and Teams chats are fair game for e-Discovery; now, the prompts feeding your forensic analysis may be, too. The protection no longer hinges on the tool itself, but on who created the prompt, the purpose, and the terms of service (and good luck if you're using a public LLM). If your incident response plan involves uploading sensitive data to an LLM to generate a report, you are effectively creating a future discovery trail.

The court also made it clear that generic Rule 29 agreements won't save you. The parties had a stipulation protecting "expert notes, drafts, or communications," but the judge ruled it wasn't "quite clear" enough to cover AI prompts. Silence on AI in your discovery procedures is now a potential liability.

You need to audit your IR playbooks immediately. If your team uses AI to analyze logs or draft findings, those inputs are likely discoverable. You must update engagement letters to explicitly address AI use, implement strict preservation protocols for prompt logs before they get purged by the platform, and be straightforward in your declarations. Trying to relabel prompts as "search terms" after the fact is a losing strategy that invites sanctions under Rule 37(b). In the eyes of the court, if the AI helped you find the needle in the haystack, the instructions you gave the machine are considered to be a part of the search.