Overview
As artificial intelligence (AI) transforms the legal landscape, it looks to be a promising tool to improve the speed, and thereby lessen the cost of otherwise time-consuming and expensive intellectual property litigation tasks. However, to effectively harness the power of AI, litigants must balance the advantages of AI with its limitations. Anticipating potential pitfalls is a first step; utilizing human-in-loop protocols to ensure responsible use of AI is not only an important but necessary second one.
Responsible Use of AI in Formulating Case Theory and Drafting
It may be tempting to use AI to assist with case theory or preparation of litigation papers, but litigants must recognize the substantive and procedural limitations of AI tools. Substantively, AI “hallucinations,” namely, fabricated or misstated information presented by AI systems as fact or law, create real problems with significant consequences. For example, in a patent litigation, an attorney personally faced covering a $10,000 repayment fee to opposing counsel and possible Rule 11 Sanctions after submitting an error-filled claim construction chart almost entirely generated by AI. While the attorney used a second AI system to cross-check the work, the second tool failed to identify the erroneous citations. Magpul Industries Corp. v. Mission First Tactical, LLC, No. 24-5551-KSM (E.D. Pa. 2025). In other cases across the United States, attorneys have faced even larger fines, sanctions, and other penalties for using AI tools and submitting fabricated case citations and inaccurate information without proper verification.
Critically, AI systems lack the contextual judgment necessary to prepare sound pleadings without human oversight, and they may not be trained properly in the various rules applicable to a case. For example, they may omit foundational facts relevant to standing or jurisdiction of the presiding tribunal, or key elements of a cause of action, especially as they relate to IP disputes. Such deficiencies may render pleadings drafted with AI technology vulnerable to summary dismissal and the filing party subject to sanctions.
Anticipating Potential Pitfalls in Discovery Where AI Tools are Concerned
With respect to discovery, litigants must be aware that materials generated by AI may be discoverable unless a privilege applies. For example, in the recent copyright litigation, Concord Music Grp., Inc. v. Anthropic PBC, Anthropic attempted to obtain broad discovery of all AI prompts and outputs used by Concord, arguing that such information was not shielded from discovery. The court disagreed and held that the attorney work-product doctrine applied and had not been waived. No. 24-CV-03811-EKL (SVK), 2025 WL 1482734 (N.D. Cal. May 23, 2025).
However, in another copyright case, the court initially ordered OpenAI to preserve all ChatGPT output logs—an extremely extensive data retention requirement. In re OpenAI, Inc. Copyright Infringement Litigation, No. 23-CV-11195, 2025 WL 2691297 (S.D.N.Y. Sept. 19, 2025). After months of negotiation, the parties ultimately agreed to a narrower preservation scope.
Also relevant in discovery is ensuring protocols are used to avoid inadvertent disclosure of confidential client information. For example, in a patent litigation, the parties agreed that any AI tools used to review or analyze discovery materials must be “fully containerized,” meaning they could not share, train on, or otherwise reuse litigation data. Eireog Innovations Ltd. v. Amazon Web Services, Inc., No. 1:25-cv-00552-ADA (W.D. Tex. Oct. 7, 2025), Dkt. 41, ¶30.
Professionals
Vera Glonina
Associate
Audrey Trace
Associate
