AI Chats and Discovery – Beware the Glass House

By Matt Wood

Businesses of all sizes use AI to move faster and do more. AI chats range across finances, strategy, and yes – even law. Some risks are well-known, like hallucinations. That’s why important information (especially legal) should be checked by a qualified human. 

But what about the confidentiality of AI chats? What happens if you or your company is sued? Courts are beginning to answer that question. And the message is clear: many AI chats will not be protected from disclosure in a lawsuit – and could end up in the hands of someone suing you.  

Every lawsuit has a process called “discovery.” The parties request and share relevant information. Discovery requests commonly ask for emails, text messages, and internal data. But discovery cannot reach privileged information. (The most common privileges protect attorney-client communications and work product for the lawsuit.) So what about those AI chats – can a hostile party in litigation request and get copies of those? 

They might. Protection falls into a few buckets:

High Risk – Chats In The Regular Course Of Business

Everyday chats with AI – when no lawsuit is brewing, and no lawyer is involved – are vulnerable to discovery. Courts are likely to treat them like other business communications, such as emails or text messages. Don’t say anything you wouldn’t want read by someone in a lawsuit. 

Medium Risk – Chats To Prepare For Litigation

When a lawsuit is threatened, non-lawyers are increasingly using AI to evaluate legal risks. Are those chats protected from disclosure, perhaps as work product? Courts don’t all agree. Some have said work-product protection can apply in this situation. [1] At least one judge has disagreed, requiring the chats to be disclosed. [2] More might follow. This area of the law is evolving. Tread lightly. 

Low Risk – Chats By Legal Counsel

Chats by lawyers stand on different ground. Work product is privileged in large part to protect lawyer analysis from disclosure. The emerging court decisions in this area have ruled that lawyers’ AI chats are protected work product. [3]

Waiver Warning!

Remember that privilege can be waived by disclosure. In one case, a client waived the privilege for his attorney’s advice by putting it into an AI chat that was not confidential. (The AI terms and conditions did not prohibit model training or disclosure to third parties.) [4] You should discuss with your attorney before putting any legal advice into an AI chat to ensure privilege isn’t lost. And watch out for employees using unapproved AI tools (so-called “Shadow AI”). Productivity may go up, but confidentiality and legal privileges could be lost. Monitor and manage employees’ use of public AI tools.

The main takeaway from this evolving area of the law is simple: be careful what you and your employees say in AI chats. What feels confidential and private may not be. When in doubt, check with your lawyer. You may be in a glass house.

[1] Morgan v. V2X, Inc., No. 25-cv-01991, 2026 WL 864223 (D. Colo. Mar. 3, 2026); Warner v. Gilbarco, Inc., No. 2:24-cv-12333, 2026 WL 373043 (E.D. Mich. Feb. 10, 2026).

[2] United States v. Heppner, No. 25 Cr. 503, 2026 WL 436479 (S.D.N.Y. Feb. 17, 2026).

[3] Concord Music Group, Inc. v. Anthropic PBC, No. 24-cv-03811, 2025 WL 1482734 (N.D. Cal. May 23, 2025); Tremblay v. OpenAI, Inc., No. 23-cv-03223, 2024 WL 3748003 (N.D. Cal. Aug. 8, 2024).

[4] Heppner, 2026 WL 436479, at *3 n.3.

About the Author(s)

Matt C. Wood

Matt C. Wood is a Partner at Vela Wood.  He focuses on complex litigation and high-stakes appeals.

Learn More