Are GenAI prompts in eDiscovery protected work product or not? Two notable experts in our industry tackle that question in their latest paper.
The 26-page paper – authored by Tara Emory and Maura R. Grossman is titled GenAI Prompts in eDiscovery: Protected Work Product or Not? and is available here for download.
As it will be evident by the second paragraph of the first page Abstract what the authors think, I’ll just go ahead and tell you what they think: “absent party agreement or a demonstrated deficiency in the production (or other discovery misconduct), prompts should ordinarily not be subject to compelled disclosure, and that prompts reflecting strategic judgment or case theory should be protected as opinion work product.”
The core of the paper is the “why” GenAI prompts in eDiscovery should be protected work product. Emory and Grossman illustrate that in several ways within the paper, including:
- Avoiding the Return of the “TAR Tax”: With original technology assisted review (TAR), responding parties often “eschewed the use of TAR because of the extensive disclosure demands placed on them (sometimes referred to as the ‘TAR tax)”. Required disclosure of prompts could cause parties to avoid using GenAI too.
- Transparency Can Also Reduce Effectiveness: Even for parties who decide to use GenAI, the routine disclosure of prompts could be counterproductive, as it could incentivize “generic” prompting that reduces search effectiveness (lower recall and precision).
- Validation Over Disclosure: Transparency is better served through outcome-based validation – such as statistical sampling, recall and precision metrics, and “null set” analysis – rather than the exposure of iterative search processes.
- Early Court Rulings Agree: Recent rulings, including Tremblay v. OpenAI and Concord Music Group v. Anthropic, point to a judicial trend toward treating attorney-crafted prompts as protected opinion work product.
As they note in the Introduction: “The authors’ position is that, while parties should cooperate and agree to disclosure of prompts when doing so will facilitate and benefit the eDiscovery process, disclosure should not ordinarily be required except to address material eDiscovery deficiencies; even then (except in cases of outright misconduct), prompts reflecting strategic thinking, mental impressions, or case theory should generally be protected as opinion work product.”
Grossman and Emory also note that while “courts have consistently found that search terms and a general description of the search methodology to be used are not protected under the work-product doctrine”, “GenAI TAR prompts…may align more closely with protected document compilations than with unprotected search terms” and that “Prompts can be more revealing than simple search terms”, potentially reflecting “attorney guidance that reflects legal theories or investigative strategy”.
Perhaps the most instructive part of the paper is the hypothetical healthcare fraud scenario of Regional Health Systems vs. Medi Staff, which starts on page 20. It walks through the prompt iteration process, as follows:
- Starting Generic Prompt: An initial review prompt from an RFP submitted by Plaintiff, which is essentially generic.
- Prompt Evolution through Fact Work Product: After witness interviews, attorneys may identify specific individuals (e.g., “Marcus Chen”) who raised internal concerns. Adding instructions to flag “expressing support for, empathy toward, or acknowledgment of Marcus Chen or Marcus Chen’s challenges of management” reveals facts learned through investigation that the opposing party wouldn’t have to spend their own effort to obtain if the prompts were disclosed.
- Prompt Evolution through Opinion Work Product: Attorneys could then add instructions to find “communications from executive leadership concerning financial performance, margin improvement, or operational efficiency targets”. This could reflect a legal theory that financial pressure is a motive for fraud, and disclosing this prompt reveals the mental impressions of the attorney regarding how to prove the case.
- Prompt Evolution to Improve Precision: To reduce false positives, attorneys might add exclusionary prompts (e.g., “Communications concerning Bio Flex Solutions’ service delays, financial instability” are not responsive). Disclosing prompts could reveal sensitive, non-relevant business information about third-parties that is outside the scope of discovery.
It’s a terrific walk through that illustrates the attorney mental impressions that are part of prompt iteration.
As part of their Conclusion section, Emory and Grossman state this:
“The authors strongly believe that the adoption of TAR was regrettably impeded by protracted battles that ensued over the disclosure of seed sets, training examples, process considerations, and examples of non-responsive documents contained in null sets. The industry should learn from its prior mistakes rather than repeat them with GenAI TAR.”
Undoubtedly, there will still be people in our industry that want to equate GenAI prompts in eDiscovery with search terms in eDiscovery, but this paper from Maura R. Grossman and Tara Emory goes a long way in refuting those arguments. Check out the paper here!
So, what do you think? Are GenAI prompts in eDiscovery protected work product or not? If you think not, why not? Please share any comments you might have or if you’d like to know more about a particular topic.
Image created using DALL-E 3, using the term “robot guard standing in front of a robot lawyer wearing a suit sitting at a workstation using ChatGPT”.
Disclaimer: The views represented herein are exclusively the views of the author, and do not necessarily represent the views held by my employer, my partners or my clients. eDiscovery Today is made available solely for educational purposes to provide general information about general eDiscovery principles and not to provide specific legal advice applicable to any particular circumstance. eDiscovery Today should not be used as a substitute for competent legal advice from a lawyer you have retained and who has agreed to represent you.