News & Insights
Hillary Hurst for the Journal Record - Generative AI is the New Diary: Risks of the Digital Confessional in Litigation
September 2026
By: Hillary N. Hurst
Journal Record
The Oklahoma rules of civil procedure permit litigants to request the production of any non-privileged documents in discovery from their adversary that are “relevant to any party’s claim or defense, reasonably calculated to lead to the discovery of admissible evidence and proportional to the needs of the case. . . .” 12 O.S. § 3226(B)(1)(a). Requesting the production of an opposing party’s personal notes or diaries during discovery has been standard operating procedure in civil litigation for years. Because a person writing in a diary or keeping personal notes has little expectation that a third-party would ever read the same, entries in diaries relating to litigation have traditionally been highly sought after as they may may reveal key facts, mental impressions, timelines, or even admissions that may otherwise be filtered, polished, or rehearsed—or omitted entirely.
From the business owner to the pro se litigant, public artificial intelligence platforms such as ChatGPT, Anthropic Claude, and Google Gemini have become a sounding board for users to evaluate risk or vet legal options. However, use of generative artificial intelligence for this purpose, including the prompts given and the results generated, may be discoverable as a modern-day diary.
Oklahoma courts have yet to directly address the issue. Courts in other jurisdictions, however, have determined that communications between a litigant and artificial intelligence are not shielded from discovery by the attorney-client privilege or the work-product doctrine. United States v. Heppner, 820 F. Supp. 3d 292 (S.D.N.Y. 2026).
As the role of generative artificial intelligence increases in daily operations, so too does the risk that what a user may perceive as a confidential communication becomes a public court record. Parties to a present or anticipated dispute who share their anxieties, legal strategies, or liability assessments with generative artificial intelligence may unknowingly generate additional legal exposure or foreclose legal avenues.
The legal landscape around this issue continues to evolve with the technology. For now, the guiding principle is simple: If you do not want information to be read aloud in a deposition or in open court, do not share it with a public generative artificial intelligence platform. In general, all privileges which may shield documents from production in discovery require, “a trusting human relationship,” which is simply absent with artificial intelligence.
