The construction industry has been quick to integrate AI into its business practices and efficiencies are being recognized. However, such adoption is not without risk. Two recent court rulings help to illustrate one risk stemming from the use of AI: the potential for waiver of the attorney-client privilege or the attorney work-product protection. The law is not settled on these issues – it is still developing. As such, it is critical that AI users be cautious and consult an attorney before inputting information relating to legal issues into AI systems or generating documents for use in litigation, mediation, or arbitration.

What Are the Attorney-Client Privilege and Work-Product Protections?

The attorney-client privilege. Communications between attorneys and their clients, for the purpose of obtaining or dispensing legal advice, are generally protected from disclosure to third parties (such as an opposing party in litigation). Importantly, this “privilege” applies only when the communications are confidential.  That is, the communications are not shared with any third party.  If an otherwise privileged communication is shared with a third party (i.e., someone beyond the attorney or their client), the privilege is usually considered waived, meaning third parties can then access the communication.

The attorney work-product doctrine. This doctrine protects the mental impressions and strategies developed by attorneys from being disclosed to an adverse party. This protection covers work-product generated by attorneys and their legal teams, and information prepared at an attorney’s direction, including documents created by an attorney’s clients.  As with the attorney-client privilege, the work-product protection can be waived if the work-product is not kept confidential.

How Does the Use of AI Affect These Protections?

Two recent decisions highlight the main concerns relating to AI and these protections, as well as the varied approaches courts take when addressing these emerging questions. In short, it is a best practice that an attorney be consulted before AI is used to weigh in on a legal issue.

In United States v. Heppner, a criminal defendant input information into Anthropic’s Claude following receipt of a grand jury subpoena. —F.Supp.3d—- (S.D.N.Y. Feb. 17, 2026). The defendant’s goal was to prepare “reports that outlined defense strategy.” Id. at *1. When these reports were seized, the defendant argued that they were protected by the attorney-client privilege and the work-product doctrine because they were prepared using information provided by his lawyers, were created for the purpose of obtaining legal advice, and were subsequently shared with counsel. Id. at *2. The court rejected the defendant’s arguments and found that the documents were not privileged “because Claude is not an attorney” and thus it cannot dispense legal advice. Id. The court further found that the information submitted to Claude was not confidential for purposes of the attorney-client privilege because the information was shared outside the attorney-client relationship, nor was it shared with Claude at the direction of legal counsel. Id. Because the information was not kept confidential, it was not privileged and thus subject to disclosure. Id.

On the other end of the spectrum is the decision in Warner v. Gilbarco, Inc. There, the court denied a motion to compel production of “all documents and information concerning [the plaintiff’s] use of third-party AI tools in connection with this lawsuit.” 2026 WL 373043 at *4 (S.D. Mich. Feb. 10, 2026). Unlike in Heppner, the court protected the information input into ChatGPT under the work-product doctrine.  The court reasoned that ChatGPT and other generative AI programs are “tools, not persons” and thus disclosing information to the programs does not waive any privilege or protection. Id. The court firmly rebuked the defendant’s attempt to access the plaintiff’s “mental impressions” through what it deemed to be an “intrusive . . . fishing expedition.” Id.

The varying outcomes in Heppner and Warner demonstrate the developing nature of this issue and the need for caution when employing AI to assist with legal issues. As stated above—and it is worth reiterating here—an attorney should be consulted before information is submitted to AI to assist with a legal issue, especially when litigation is imminent or already pending.

The Hahn Loeser Construction Team is monitoring court decisions surrounding the use of generative AI, implementing AI processes within our firm, and collaborating with clients to develop internal AI systems and policies. We are available to answer questions about using generative AI in construction and other industries to develop best practices for protecting your confidential information.