Abstract
When a layperson uses ChatGPT to obtain feedback on a legal matter, attorney-client privilege may not apply, as ChatGPT is not a lawyer, much less a human. Further, while lawyers are entitled to protection for their opinion work-product, it is not clear whether self-represented litigants are entitled to the same protection. Additionally, the broader duty of confidentiality binds only attorneys, not AI systems like ChatGPT. The public increasingly uses AI tools such as ChatGPT. If a layperson employs ChatGPT for legal advice, particularly in a civil matter, such communications may be discoverable and potentially admissible. This presents an access to justice issue because a self-represented litigant who seeks to understand the scope of their legal rights may not realize that their AI communications can be used against them. Alternatively, they may not be able to afford an attorney, and thus decide to take the risk of communicating with AI anyway. This Article argues that self-represented litigants should enjoy protection for opinion work-product, and further, AI responses to self-represented litigants should also be permitted to count as opinion work-product. In addition, this Article proposes a discovery management protocol so courts may handle AI communications in a practicable manner. The work-product solution may be implemented more easily than other options, such as extending attorney-client privilege, though that also may be advisable in the interests of justice. Finally, as a backstop, judges could also consider excluding such AI communications from admission under Rule 403.
Citation
Anoo D Vyas, Confidentiality of AI Conversations: Protecting Self-Represented Litigants Who Use ChatGPT for Legal Advice, 26 Duke Law & Technology Review 186-232 (2026)
Available at: https://scholarship.law.duke.edu/dltr/vol26/iss1/7