Courts are increasingly faced with deciding whether information litigants share with public artificial intelligence (“AI”) services and the AI’s output (together, “AI materials”) can be attorney-client privileged or protected by the work product doctrine. Two recent decisions from the Eastern District of Michigan and Southern District of New York have reached opposite conclusions on the issue, highlighting that litigants and employers will increasingly need to take care to avoid potentially waiving privilege or work product protections through their use of AI services.
First, in the Eastern District of Michigan action, the court held that the work product doctrine protected materials a pro se plaintiff generated using a public AI service. The plaintiff used the AI service to upload case documents, draft filings and answer legal questions. The court held the AI materials protected by the work-product doctrines, characterizing them as the plaintiff’s internal analysis and mental impressions. Critically, the court found the use of the AI service did not waive work-product protections because, while using the AI service disclosed the contents to it and, potentially, its administrators, that did not disclose the materials to an adversary or in a way likely to get in an adversary’s hands.
Conversely, in the Southern District of New York case, the court held a defendant with legal counsel cannot claim privilege over chat exchanges between himself and a public AI service regarding a pending criminal investigation, even where he input information received from counsel and where he shared the AI-generated results with his attorneys for discussion. The court emphasized that, to the extent the AI materials contained information the defendant learned from his legal team in the course of their representation, the defendant waived attorney-client privilege by disclosing it to the third-party AI provider, whose terms expressly provided that any material submitted was not confidential.
The Southern District of New York opinion held the AI materials not privileged for reasons that will generally be applicable to discussions with many, if not all, AI platforms:
- Not communications between a client and their attorney. The privilege protects communications with lawyers—not communications with non‑lawyers such as an AI tool, even if a client finds the AI-generated content helpful for subsequent discussions with counsel.
- Not confidential. AI materials will generally not be confidential because commercial AI platforms’ privacy policies permit the collection, training use and disclosure (including to governmental authorities and third parties) of the inputs users provide and the output the AI services generate. As personal as AI conversations may feel, users should bear in mind that the AI provider may disclose the information provided to the fullest extent allowed under their privacy policy.
- Not for the purpose of obtaining legal advice. It cannot be said that AI materials are created for the purpose of obtaining legal advice from counsel where the AI tool’s terms disclaim such purpose. For example, under OpenAI’s terms of use, users agree they will not rely on its output “as a substitute for professional advice.”1 However, the Southern District of New York court expressly reserved for another day the question of whether privilege could reach AI materials the client created at his counsel’s express direction, in which case might be said the AI platform was performing a function akin to a highly trained professional acting as a lawyer’s agent, within the protection of the attorney-client privilege.
Second, the Southern District of New York court also found that the work product doctrine did not extend to the AI materials. Unlike in the Eastern District of Michigan case, they were prepared not by counsel or at counsel’s direction but rather by a defendant with legal representation on his own initiative.
Revolutionary as AI tools may be, both opinions were based on the application of longstanding legal principles, and their reasoning is applicable generally to civil as well as criminal matters.
Key Implications for Defendants and Corporate Counsel
These opinions signal a new field of potentially damaging privilege waivers in the dawning AI age. As people grow increasingly accustomed to posing important questions of AI tools, defendants and employees will surely be tempted to use AI to help them understand the risks they or their employers face. But giving in to such temptation risks producing unprivileged and discoverable documents.
To get ahead of these risks, companies and their counsel should consider the following best practices:
- First, companies should consider revising their legal, compliance, internal investigations and incident-response policies to prohibit employees from posing legal questions or prompts containing matter-relevant facts to any AI tool without approval from in-house counsel.
- Second, companies should consider whether to add content to their regular training programs teaching their employees not to input sensitive information in AI prompts.
- Third, in-house counsel should brief key personnel involved in investigations or litigations on the importance of not using AI tools, for example, at the same time as they distribute hold notices or brief them on document retention obligations.
- Fourth, counsel to parties in or facing litigation should warn clients not to use AI tools, including inputting legal questions, strategy or advice from counsel, and should discuss any AI use with counsel before acting.
- Fifth, to the extent that AI tools may be used in connection with legal work, such usage must be routed through the company’s internal or external counsel and restricted to AI tools that provident stringent confidentiality protections such as Harvey AI, Lexis+ AI or Thomson Reuters Cocounsel.
- Finally, if employees or defendants mistakenly create non-privileged AI content regarding a sensitive legal issue, they should not compound the error by emailing the content around and should notify counsel immediately. Sharing AI-generated content with legal counsel for discussion will not extend the privilege back to the AI materials and will generate additional documents or messages potentially alerting the government to their existence.
This notice provides general guidance on preparing for the new challenges AI technology is bringing to discovery in criminal and civil matters. For more information, please contact a member of Haynes Boone’s White Collar and Investigations Practice Group.
