The issue of privilege related to a client’s use of AI came storming into the discussion due to two cases decided on the same day, Heppner and Warner.
Cases Beginning in March 2026
As you may recall, Heppner held that if a criminal defendant is represented by counsel, their discussions with AI is not privileged and may be discovered. Heppner was decided based upon the privacy policy of the AI tool the defendant used. Warner held that for a pro se party, use of AI is work product, and protected. Heppner left open what would happen if a lawyer directed the client to use the AI. Many saw these cases in opposition, but given the fundamental difference involved whether the person was pro se, it was hard to draw conclusions.
Similarly to Warner, Morgan held that a pro se individual’s use of AI is protected as work product. Further, Morgan specifically noted that if we allowed use of a tool to void privilege simply because it was cloud-based, we would have a problem due to how many cloud-based services attorneys, clients, and pro se individuals use. But fundamentally, Morgan still turned on the notion that a pro se individual is protected through work product, since they are acting as their own attorney. To my knowledge, no one has successfully argued that when a lawyer uses AI that vitiates attorney/client privilege, regardless of the privacy policy.
A case I focus on that many others do not is Fortis. A Delaware Chancery Court opinion, Fortis is interesting for what the court does not mention. At no point does the court analyze whether AI chats should be protected. The court simply allows the admissibility of the data.
There are many unreported cases like Fortis in which the assumption is that AI chats are discoverable. When I speak with family law attorneys, they tell me that courts are frequently admitting AI chats in their cases. There are no authored opinions that I have seen in this regard, but family law courts are often on the cutting edge of discovery and technology.
The Newest Case: Assini v. Hayward
The next important case in the analysis of discovery and AI is Assini v. Hayward. I have not previously written about Assini, so I will take a moment to provide an in-depth exploration of the opinion. Assini is another pro se case. As always, the pro se posture matters for the final conclusion about discoverability of AI chats. Essentially Plaintiff sought a substantial amount of information from OpenAI related to Defendant’s use of ChatGPT. Defendant, of course, objected and sought to prevent or limit that access.
In its analysis, the court first noted that “Materials prepared in anticipation of litigation are subject to a conditional privilege (CPLR § 3101[d]). To demonstrate that the anticipation of litigation privilege under CPLR § 3101(d) is applicable, it must be shown that ‘the material [was] prepared solely in anticipation of litigation.'” Assuming Defendant could successfully convince the court that a conditional privilege was appropriate, Plaintiff would need to show that there was a “‘substantial need’ and [would be] ‘unable to duplicate’ [the requested materials] without undue hardship.”
Unsurprisingly, Plaintiff turned to Heppner, which, given the pro se nature of the instant case, was not particularly helpful to their argument. Rather, the court cited Morgan, noting that, “‘the use of AI closely resembles the kind of confidential, strategy-laden iterative work product that Rule 26(b)(3) was designed to protect.'” The court noted that it “finds the rationale presented in Morgan persuasive” and quashed Plaintiff’s subpoena.
Interestingly, the court also stated, “defendant Recchio’s use of AI has not gone unnoticed by the Court. While the Court Rules do not prohibit the use of AI, the defendant’s ‘use of AI frustrates the litigation’ and cannot go unfettered.” The court further instructed Defendant to review “22 NYCRR §161, ‘Use of Artificial Intelligence Technology’ and directed him to be governed accordingly.” The court also warned Defendant about potential sanctions for misuse of AI. This is a new rule in New York and only recently went into effect.
What Do We Know About AI and Discovery?
Currently, if an individual is pro se, their use of AI in preparation for litigation is likely to be protected as work product. If an individual is represented and was not directed to use AI by their attorney, their AI chats are unlikely to be protected. As noted, Heppner left open whether AI chats are discoverable if an attorney directs that use. Until a court gives us guidance on this issue, the answer is we do not know. But privilege limited to instruction by counsel is a logical next step. There is also a question as to whether the Warner and Morgan courts would extend protection beyond pro se individuals, especially the Morgan court, which brought up issues related to cloud-based software in general.
How to Protect Your Clients
If clients are not aware of the potential that their AI conversations could be discovered, they could be quite embarrassed, and if they write the wrong thing, such content could conceivably destroy their cases. The best way to protect your clients from having their AI chats discovered is to warn them about its use. I recommend that this warning be in writing, separate from the engagement agreement. Have your clients sign the warning. That way they cannot later claim you failed to warn them against using AI.
Sample Clause
Here is a clause you might consider adapting to your own engagement agreements. Attorneys may use it freely. Please remember that using this clause does not form an attorney/client relationship between you and me or Jennifer Ellis, JD, LLC.
A Word of Caution About AI Tools. Please do not use AI tools like Claude, Copilot, or ChatGPT to discuss your case or to write messages about your case. When you type something into an AI tool, that information often gets stored on the company’s servers, and people at the company may be able to read it. Your conversations with an AI tool are not private the way your conversations with us are. In addition, the other side may be able to demand copies of what you typed into these tools, and a judge may order you to hand them over. This could hurt your case. The protection that keeps your communications with your lawyer confidential does not cover what you share with an AI company. If you have questions about your case, ask us directly. Do not ask an AI tool.
Note: You might also consider requiring that your clients reveal AI use to you.