Confidentiality, Work Product, and Privilege in the Age of AI: Lessons From Early Court Decisions + Practical Employer Takeaways

Confidentiality, Work Product, and Privilege in the Age of AI: Lessons From Early Court Decisions + Practical Employer Takeaways

More courts are confronting modern issues related to confidentiality, work product, and privilege as an increasing number of litigants and law firms are using artificial intelligence for assistance with lawsuits. This Insight provides both snapshot and in-depth looks at some of the earliest court decisions to addresses these AI issues – and offers three practical takeaways for employers.

Deeper Dives

Two early courts split on protections for AI use in litigation.

• Warner v. Gilbarco, Inc. In an employment discrimination suit, a federal court in Michigan ruled in February that the work-product doctrine protected a pro se plaintiff’s ChatGPT materials used in connection with the lawsuit.
• United States v. Heppner. One week later, a federal court in New York held that a criminal defendant’s communications with Claude for legal advice (which he engaged in of his own volition and did not reflect his counsel’s legal strategy) were not protected by the attorney-client privilege or the work-product doctrine – demonstrating how disputes may play out differently across jurisdictions and depending on the specific facts of each case.

Read more about these two cases here: Can Your AI Chat History Be Used Against You in a Lawsuit? 5 Practical Takeaways for Employers as Courts Start to Split.

Since then, more courts have chimed in – and new issues have emerged.

Morgan v. V2X, Inc.

In a second employment discrimination case to address litigation protections for AI use, a federal court in Colorado ruled in March that some level of work-product protection applied to a pro se plaintiff’s use of AI in preparation for the litigation. Specifically, the court determined that the outputs from the AI system were protected, but not the name of the AI tool used. The court said that the employer needed the name of the AI tool to assess whether the plaintiff had compromised certain confidential information by submitting it into a mainstream AI tool, and that the plaintiff failed to show that revealing the tool’s name would reveal his mental impressions or legal strategy.

The court also addressed the use of mainstream AI tools to process confidential information. Ultimately, the court amended the parties’ protective order – which prohibited each party from disclosing certain confidential information of the other party involved during the discovery process – to prohibit the plaintiff from uploading, inputting, or submitting confidential information into any AI platform unless the AI provider is bound by certain contractual safeguards.

While the court recognized that this type of restriction disadvantages pro se litigants who do not have access to or who cannot afford enterprise-AI tier accounts that can satisfy the confidentiality requirements, it said it could not “ignore the real risks associated with mainstream tools that persistently collect and store data and could compromise confidentiality.” The court also made it clear that “modern AI tools may be used in many ways that do not involve uploading confidential information.”

Jeffries v. Hacros Chemicals Inc.

Just days before Morgan decision above, a federal court in Kansas made a similar AI-related amendment to a protective order governing discovery in the case (which is not related to employment law or the workplace). However, this court took it a step further by prohibiting the parties from uploading even non-confidential documents and materials produced by another party into public or “open loop” generative AI tools.

The court said that while using generative AI tools for eDiscovery tasks can help save significant time, such use should be limited to so-called “closed” AI tools, as using open AI tools for this purpose could expose massive amounts of data and potentially violate US data privacy laws and the EU’s stricter General Data Protection Regulation. The court also said that allowing parties to use open AI tools in this context could lead parties to “err on the side of under-producing potentially responsive documents or seek to make extensive redactions of irrelevant or non-responsive information.”

Shealy v. Seaside Investments, LLC

A Massachusetts court issued a decision on June 16 that determined, as a matter of first impression, that the work-product doctrine under the state’s Rules of Civil Procedure (unlike the cases above, which all involved federal rules) did not apply to AI queries or output for legal assistance created by a plaintiff and his romantic partner in a case involving alleged breaches of a loan and note purchase agreement.

Further, the court made it clear that the output from ChatGPT does not constitute opinion work product: “Al is a tool not a person. Nothing about what ChatGPT produces discloses the "mental impressions, conclusions, opinions, or legal theories of an attorney or other representative of a party concerning the litigation.”

The court also said that this case was not comparable to Morgan and Warner, and was more akin to Heppner, since the plaintiffs were represented by sophisticated, capable counsel who did not direct the plaintiffs (or their romantic partners) to use ChatGPT to generate information and documents.

3 Key Takeaways for Employers

As more courts confront protections related to AI use in litigation, a few patterns are starting to emerge:

• Courts are more likely to apply work-product doctrine to AI materials generated by pro se plaintiffs and less likely to do so when a party is represented by counsel and uses AI for legal assistance that is unrelated to their counsel’s strategy or advice. You should therefore tread carefully when using AI for legal advice and train your leadership on the same.
• Courts are more likely to enter protective orders during the discovery phase that prohibit parties from uploading confidential information – or even non-confidential information – into open AI tools or AI platforms that lack certain contractual safeguards. You should therefore be prepared for how this could impact discovery tasks and timelines. You should also vet your AI vendor before deploying their technology.
• Federal and state rules regarding the work-product doctrine and other civil procedure issues can vary, potentially making it harder to qualify for certain protections for AI-generated materials in certain jurisdictions. Work with counsel to understand the rules that apply to your business in each case.

However, this is just the beginning. Stay tuned for updates as more courts across the country weigh in and shape this new area of the law in various contexts and circumstances.

Conclusion

We will continue to monitor developments in this area and provide updates as warranted, so make sure you are subscribed to Fisher Phillips’ Insight System to get the most up-to-date information directly to your inbox. If you have questions, please contact your Fisher Phillips attorney, the authors of this Insight, or any member of our Litigation Practice Group or our AI, Data, and Analytics Practice Group.