Your AI Chat Logs May Be Exhibit A
Category | Briefing Papers
By Jesse R. Orman
Jesse is a shareholder of the firm’s Construction Law Department. He can be reached at 612.359.7637 or jorman@fwhtlaw.com.
Generative artificial intelligence has quickly become part of the construction industry’s daily workflow. Contractors are using AI tools to summarize project records, analyze scheduling impacts, evaluate potential delay claims, draft correspondence, review contract provisions, and even develop preliminary litigation strategies.
The efficiencies are obvious. A project executive can upload thousands of pages of project documentation and ask an AI system to identify potential change-order issues. A project manager can ask an AI tool to evaluate whether a subcontractor’s notice complies with contractual requirements. A claims consultant can use an AI platform to summarize competing schedule analyses. And when disputes arise, personnel often use generative AI to organize facts, test arguments, and evaluate potential claims before consulting counsel. What not so long ago felt like an experimental technology has become an increasingly common business tool.
Many construction professionals, however, may be operating under a dangerous misconception: that interactions with AI tools somehow disappear into a digital black box. They do not. As courts begin to confront the discoverability of generative AI materials, the small but growing body of authority suggests that prompts, chat logs, uploaded materials, generated reports, and related metadata may be treated like any other electronically stored information (“ESI”). This article explores the implications of this development for the construction industry.
1. Are AI Materials Just Another Form of ESI?
For more than two decades, litigants have understood that emails, text messages, instant messages, Teams chats, electronic drafts, and essentially all other digital records are potentially discoverable in litigation. For nearly as long, litigants have understood the implications of not preserving those records. Organizations have developed sophisticated retention policies, litigation hold procedures, and ESI preservation protocols to address those realities.
Generative AI usage creates another category of electronically stored information. When a user interacts with an AI platform, numerous records may be created, including prompts and questions, uploaded project files, AI responses and outputs, follow-up conversations, generated drafts, user and date metadata, and internal logs maintained by the AI platform provider.
From a discovery perspective, these materials often look remarkably similar to emails, internal memoranda, draft reports, or Teams communications. Courts therefore have little conceptual difficulty treating them as discoverable ESI.
Construction companies should not assume that because AI-generated analyses appear in a conversational interface, rather than a traditional document, they are somehow insulated from discovery. The format has changed; the discovery principles have not.
2. The Emerging Case Law
Two decisions issued in early 2026 have become focal points in the developing conversation regarding AI-generated materials and discovery obligations.
In United States v. Heppner, 820 F. Supp. 3d 292 (S.D.N.Y. 2026), the United States District Court for the Southern District of New York addressed whether documents generated through a publicly available AI platform were protected by attorney-client privilege or the work-product doctrine. According to the court, the defendant used a generative AI system to develop analyses relating to a government investigation and potential defense strategies. The court ultimately concluded that the materials were not protected by either the attorney-client privilege or work-product doctrine and ordered disclosure. The court’s reasoning was straightforward. The attorney client privilege did not apply because: (1) the AI materials were not communications between the defendant and his attorney; (2) disclosure to a third-party platform with no reasonable expectation of confidentiality destroyed any privilege that might otherwise have attached; and (3) defendant did not communicate with AI tool for the purpose of obtaining legal advice. Similarly, the work-product doctrine did not apply because the materials were not produced at the direction of defendant’s counsel, nor did they reflect the counsel’s strategy.
Although Heppner arose in a criminal context and focused primarily on privilege issues, litigants have increasingly cited the decision for a broader proposition: interactions with generative AI systems are not automatically entitled to special protection merely because they involve legal analysis or litigation-related subject matter.
The Eastern District of Michigan reached a different result in Warner v. Gilbarco, Inc., 820 F. Supp. 3d 629 (E.D. Mich. 2026), just a week after Heppner. There, the court declined to compel discovery of certain AI-assisted litigation materials prepared by a pro se litigant and concluded that work-product protection applied under the circumstances presented. The court rejected the notion that using an AI platform necessarily waives work-product protection because AI programs are “tools, not persons.”
Both Heppner and Warner are routinely discussed in the AI cases that have continued to roll out in the first half of 2026, and in related commentary. See e.g., Assini v. Hayward, No. 607683/2024, 2026 WL 1677232 (N.Y. Sup. Ct. June 4, 2026) (holding Warner was applicable, Heppner was not, and pro se defendant’s AI use was protected work product); Morgan v. V2X, Inc., No. 25-CV-01991-SKC-MDB, 2026 WL 864223 (D. Colo. Mar. 30, 2026) (holding Warner was applicable, Heppner was not, pro se plaintiff’s AI use was protected work product, but the name of the AI platform pro se plaintiff used was not); Senior Judge Herbert B. Dixon Jr., When is a Litigant’s AI Research Discoverable?, 65 No.2 Judges’ J. 37 (2026).
While there is an obvious apparent tension between Heppner (requiring production of AI materials) and Warner (protecting AI materials from discovery), the construction industry should take away a more practical lesson. Neither court suggested that AI materials exist outside the ordinary discovery framework. Instead, both courts analyzed the materials using traditional doctrines governing discoverability, privilege, confidentiality, and work product. Accordingly, what may ultimately prove to be the most important principle emerging from the early cases is that courts appear inclined to treat AI-generated materials as simply another category of ESI subject to existing discovery rules.
3. Why This Matters in Construction Disputes
The construction industry is poised to present a uniquely fertile environment for AI discovery issues. Major construction projects generate enormous volumes of information. Daily reports, schedules, RFIs, submittals, change requests, cost records, meeting minutes, photographs, emails, text messages, and project management platform data may already create terabytes of discoverable information in a significant dispute. AI tools are now – and increasingly will be – used to organize and evaluate that information.
Consider a few common scenarios involving use of an AI platform or tool: a project executive uploads delay correspondence and asks whether the owner’s actions constitute a compensable delay; a claims manager provides project schedules and requests an analysis of critical path impacts; a superintendent uploads photographs and daily reports and asks whether there is sufficient evidence to support a differing site conditions claim; or a contractor’s employee asks for an assessment of weaknesses in a pending claim before meeting with outside counsel.
In each instance, the prompts themselves may reveal the user’s thinking. The uploaded materials may identify what facts the user considered important. The AI output may contain favorable or unfavorable analyses. Follow-up questions may expose perceived weaknesses in a claim or defense. The AI chat history could very well appear to be a roadmap of the organization’s internal evaluation process. Experienced construction litigators immediately recognize the significance. Opposing counsel routinely seek documents revealing contemporaneous assessments of project issues. A claim analysis generated through an AI platform may prove every bit as interesting to an adversary as an internal memorandum prepared by a project executive.
4. The Historical Parallel: Email All Over Again
Construction lawyers (like the author) and other professionals who lived through the early years of electronic discovery may experience a sense of déjà vu.
When email became commonplace in the 1990s and early 2000s, many organizations viewed it as an informal communication tool. Users frequently wrote messages they would never have committed to paper. Few organizations had mature retention policies. Litigation soon revealed the consequences.
Courts consistently treated emails as discoverable business records. Companies learned difficult lessons about document preservation, litigation holds, and employee training. Eventually, organizations developed sophisticated governance structures to manage email and other forms of ESI.
Generative AI presents a remarkably similar challenge. Many users currently view AI conversations as informal brainstorming sessions. They may not appreciate that the resulting records are potentially searchable, preservable, and discoverable. They may not understand platform-specific retention practices. They may inadvertently disclose confidential project information or proprietary business data.
The legal risks arise not because AI is unique. They arise because users often fail to recognize that AI interactions create discoverable records.
5. Looking Ahead
The law governing generative AI remains in its infancy, but courts will continue to refine how traditional doctrines apply to emerging technologies. What already appears to be a safe bet, however, is that courts are unlikely to create an entirely separate discovery regime for AI-generated materials. Early decisions such as Heppner and Warner largely apply traditional legal principles to new technological circumstances. Future decisions likely will address retention obligations, spoliation issues, privilege questions, and discovery scope in greater detail. In just a few years, this will no longer be a new or emerging issue. But it will be applied to AI materials being generated now.
For construction industry participants, the lesson is straightforward. Every prompt, every uploaded document, every generated analysis, and every AI conversation today has the potential to become part of the evidentiary record in tomorrow’s dispute.
Twenty years ago, the world learned to treat email as discoverable evidence. Today’s organizations should begin treating AI-generated materials the same way. Companies that develop thoughtful governance policies now will be far better positioned when the next major project dispute arrives–and when opposing counsel asks for the AI chat logs.
Announcements
Congratulations to the ten attorneys from Fabyanske, Westra, Hart & Thomson, P.A. who have been named 2026 “Minnesota Super Lawyers”. The polling, researching, and selecting of “Super Lawyers” is designed to identify Minnesota lawyers who have attained a high degree of peer recognition and professional achievement. Only five percent of Minnesota attorneys receive this honor. FWHT’s 2026 “Minnesota Super Lawyers” include Mark Becker, Hugh Brown, Matt Collins, Julia Douglass, Rory Duggan, Kyle Hart, Jesse Orman, Elise Radaj, Nathan Sellers and Dean Thomson. Dean Thomson was also selected as a Top 100 “Super Lawyer”.
Congratulations to the Fabyanske, Westra, Hart & Thomson, P.A. attorneys who have been named Super Lawyer’s 2026 Minnesota “Rising Stars”. They are Colin Bruns and Erinn Valine. “Rising Stars” are nominated by their peers and must be 40 years old or under, or have been practicing for 10 years or less. No more than 2.5 percent of the lawyers in the state are named to the list.
Six attorneys from Fabyanske, Westra, Hart & Thomson, P.A. have been named to the 2026 Legal 500 City Elite listing.
The selected attorneys and practice areas in which they were recognized are:
- Julia J. Douglass – Commercial Disputes
- Rory O. Duggan – Banking & Finance
- Fred H. Ladner – Banking & Finance
- Katie A. Welsch – Banking & Finance
- Paul W. Muilenberg – Corporate and M&A
- Thomas J. Tucci – Corporate and M&A
Also, Fabyanske, Westra, Hart & Thomson’s Construction Group were ranked nationally for the first time this year.