As generative artificial intelligence (AI) becomes part of everyday legal drafting and analysis, courts are beginning to confront a basic question: When does the use of a third-party AI tool threaten attorney-client privilege or work product protection? Recent decisions at the federal court level do not support a categorical rule that AI use alone waives either protection.
Instead, two recent decisions in U.S. v. Heppner and Warner v. Gilbarco reinforce a narrower point: established privilege doctrines still govern, and outcomes will depend on the nature of the material, the role of the user, and the conditions under which the tool was used. In-house teams independently using a public AI platform, an employee using an enterprise tool in the ordinary course of business, and a litigant using AI as part of a legal drafting or case-preparation process do not present the same privilege or discoverability questions. That distinction may affect not only whether a protection attaches in the first place, but also whether the resulting exchanges are later treated as ordinary business records, litigation-preparation materials, or unprotected third-party communications.
Attorney-Client Privilege and Work Product Protection 101
Attorney-client privilege protects confidential communications between attorneys and clients made for the purpose of seeking or providing legal advice. The doctrine protects the communication itself. It does not protect underlying facts, and it can be lost if the communication is shared with, or made in the presence of, third parties who are not necessary to the legal representation.
Work product protection shields materials prepared in anticipation of litigation by or for a party or its representative, including attorneys. The doctrine is aimed at protecting litigation preparation, particularly counsel’s strategy, mental impressions, and legal theories. Unlike attorney-client privilege, it is not limited to lawyer-client communications and can extend to documents prepared by nonlawyers if they were created because of litigation.
Background on Two Federal Court Cases
In U.S. v. Heppner, Judge Jed Rakoff of the U.S. District Court for the Southern District of New York held on February 17, 2026, that documents the defendant generated by querying the consumer version of Claude were not protected by attorney-client privilege or the work product doctrine. The court’s reasoning was narrow but important. The Claude exchanges were not confidential lawyer-client communications, and the materials were not created at counsel’s direction, even though the defendant later shared the outputs with his lawyers and used information learned from counsel in the prompts. The decision turned less on AI itself and more on traditional privilege principles applied to a public, third-party tool used independently by the client.
One week earlier, in Warner v. Gilbarco Inc., Magistrate Judge Anthony Patti of the U.S. District Court for the Eastern District of Michigan reached a different result on work product. The court denied a motion to compel broad discovery into a pro se plaintiff’s use of ChatGPT and other AI tools, holding that the requested materials were protected work product. The key point was that the defendants were seeking the plaintiff’s internal litigation analysis and drafting process, and the court rejected the argument that merely using a third-party AI tool automatically waives protection. In that court’s view, treating AI use alone as a waiver would effectively erase work product protection in modern drafting environments.
What the Courts Did Not Rule On
A clean way to reconcile Heppner and Warner is to recognize that neither court announced a broad anti-AI or pro-AI rule. Instead, both courts applied conventional principles to different records and different protections.
Taken together, the decisions suggest that outputs from public consumer AI tools may still receive work product protection in some circumstances, but those same interactions are far less likely to satisfy the requirements for attorney-client privilege. The key variables are who used the tool, under what conditions, for what purpose, and which doctrine is being invoked.
Heppner asked whether a client’s self-directed exchanges with a public AI platform could qualify for attorney-client privilege or work product protection. Warner asked whether the use of AI alone waives work product protection over litigation-preparation materials. One court refused to convert public AI interactions into privileged communications. The other refused to adopt a rule that AI use automatically destroys work product protections.
The Important Takeaway for Businesses
Businesses should draw a hard line between legal and non-legal uses of AI tools. Non-legal teams should not upload attorney-client communications, legal advice, draft pleadings, investigation materials, or other sensitive legal content into consumer AI tools, because disclosure to a third-party platform can create serious waiver risk for attorney-client privilege.
By contrast, an enterprise-licensed tool may offer materially better contractual, administrative, and data-handling protections, including commitments that business inputs will not be used to train models and the availability of stronger administrative controls. But those protections do not transform ordinary business chats into privileged communications. OpenAI states, for example, that consumer services may use content to improve models unless users opt out, while ChatGPT business and enterprise do not use business inputs for training by default.
Even so, employee chats with an enterprise AI tool, when those employees are not acting within a legal-team workflow, may still be discoverable in litigation like other business records. The practical takeaway is straightforward: Reserve legal-content workflows for approved, controlled environments under legal oversight, and do not assume that an enterprise license eliminates discovery risk for non-privileged AI conversations.
For more information, please contact us or your regular Parker Poe contact. Click here to subscribe to our latest alerts and insights.