Attorney Professionalism Forum: Talking to AI Could Talk You Out of Attorney-Client Privilege
8.28.2026
To the Forum:
I am a longtime New York lawyer, and a very busy one. I am writing because I have recently had to deal with the use of artificial intelligence to record conversations and meetings. AI has made it easy to create transcripts of conversations, especially on Teams or Zoom calls. All one has to do is press a button and, like magic, the result is a record of the conversation that includes transcripts and summaries that recognize voices to determine who is speaking. Some of this technology even edits out foul language. I am concerned that the use of this technology creates risks and all sorts of ethical pitfalls. I know that I cannot record conversations with clients without their consent, but what rules apply when clients record conversations with me? Do I have an obligation to discuss the use of AI and its risks, including attorney-client privilege issues? Additionally, is this something that I should address in my engagement letters?
The use of AI technology to create notes and transcripts is not limited to attorney-client conversations. I know that my clients are being told that handwritten notes are old-fashioned and that they should buy software that automatically creates transcripts of all conversations and meetings. Some of my clients are also using AI to summarize documents and meetings, which are sometimes sent to me, but I am concerned that they may become discoverable and I’m not sure whether they are privileged materials.
I recognize that the rules regarding the use of AI in the situations that I have outlined are rapidly evolving, but I would appreciate guidance from you that would put me on the right track.
Sincerely,
G. O. Skynet
Dear G. O. Skynet:
Technology is advancing at an extraordinary pace. Throughout history, innovations have propelled humanity forward: Carl Benz invented the first gasoline-powered automobile in 1886; the Wright brothers achieved the first manned flight in 1903; Robert Goddard launched the first liquid-fueled rocket in 1926; the American moon landing occurred in 1969; and the Soviets established the first space station in 1971. The world has seen the invention of Charles Babbage’s analytical engine in 1837, Lord Kelvin’s analog computer in 1872, the first digital computers in the 1940s, the Advanced Research Projects Agency Network in 1969, the World Wide Web in 1989, Google’s growth in the late 1990s and YouTube’s creation in 2005. We’ve progressed from the simple game Pong in 1972 to the immersive world of Call of Duty 4: Modern Warfare in 2007. Remarkably, the pre-microwave oven, pre-washing machine, pre-electric refrigerator, and pre-television era is still within living memory, and some individuals released from prison have never used an ATM before.[1] Today, technological change happens so rapidly that even the most visionary minds find it nearly impossible to predict what the next few years will bring. Though George Jetson was supposed to have been born in 2022, to the best of our knowledge “The Jetsons” had no laptop computers; even Captain Kirk sometimes used paper.
ChatGPT, for its part, was released in November of 2022, and the subsequent show-jumping leaps and bounds made in the performance of “artificial intelligence” have instantly transformed the very meaning of that term into the interactive tool of conversation, image and analysis generation, for which it is recognized today. Our brief historical survey demonstrates the likelihood that we are in for many surprises in the future capabilities of “generative AI.”[2] And yet, the Rules of Professional Conduct have not changed even one letter to acknowledge this new reality.
This great advance has raised and will continue to raise many questions for lawyers to consider, from the practical (Is the product of AI copyrightable?) to the ethical (Can a lawyer use AI to research law or write briefs?). As to ethical questions, fear not: It is our humble opinion that, let technology move however fast it likes, but the New York Rules of Professional Conduct constitute fundamental rules of ethical and professional conduct that are true today, were true yesterday, and hopefully will continue to be true tomorrow.
To begin, you are absolutely correct: Under New York’s ethical guidelines, even though New York is a one-party-consent state, attorneys are not permitted to record calls with a client unless the client has given consent.[3] However, the catch for lawyers is that the reverse is not true. Clients are not subject to the same rules that govern lawyers, and the use of AI or other types of tools by clients to record and analyze conversations with their attorneys is becoming increasingly common. The internet is full of ads marketing notetaking products that are literally capable of creating transcripts of virtually every telephone call, conversation, or meeting. This practice, although convenient, can create significant risks — not only for the client but also for the lawyer — and may jeopardize the attorney-client privilege by creating discoverable evidence. As experimentation with AI continues to grow, more clients, attorneys and even judges are exploring generative AI to manage the overwhelming volume of information that litigation brings,[4] amplifying these risks every day.
Although this may be a little above our pay grade, the way in which a generative AI system works is that it learns how to generate its conversations or analyses based on previous interactions with users. Therefore, AI companies usually record users’ inputs and make permission to record them a prerequisite for using the AI tool so that they can be used to train it to respond to others. People who wish to ask a chatbot about their trade secrets, without the chatbot’s relating these secrets to other users, can sometimes purchase a sort of private room that keeps all conversations between the user and the bot from being repeated to other users publicly. But even then, AI companies can still “hear” what is said in that room. The fact that the company that controls the AI system is given permission by the user to listen in has recently been found in the Southern District of New York to squarely destroy the attorney-client privilege.
In that case, United States v. Heppner,[5] a defendant, without first clearing it with his lawyer, made communications about his case to a conversant generative AI program. The court permitted disclosure of these communications because it found that “when a user communicates with a publicly available AI platform in connection with a pending criminal investigation, … the AI user’s communications [are not] protected by attorney-client privilege or the work product doctrine.” (There is no reason why this result cannot be extended to civil cases.) Of the three elements required for the attorney-client privilege to take hold, all three were lacking: (1) there were no communications between client and attorney because the program was not an attorney, (2) the communications were neither kept confidential nor intended to be kept confidential because the program warned that it collected inputs for the purpose of training the program and (3) the communications were not made for the purpose of seeking legal advice because the program was not a lawyer and stated that its output did not constitute legal advice. This broke the privilege. In addition, the work-product protection did not apply because the communications were neither “‘prepared by or at the behest of counsel’ … nor did they reflect defense counsel’s strategy.” This case highlights the danger to a client in communicating details of his case to AI: Under normal circumstances, these communications will be subject to disclosure. Heppner is not the end of the story, and to see where this all eventually ends up may require a crystal ball.
Another recent federal case, Warner v. Gilbarco, Inc., et al.,[6] ruled that “the work-product waiver has to be a waiver to an adversary or in a way likely to get in an adversary’s hand,” so that waiver of the attorney-client privilege does not necessarily waive the work-product protection where it otherwise exists. This caveat should not be misunderstood: it may keep the attorney work-product protection intact, but it does not, outside of exceptional circumstances, shield from disclosure (a) a client’s own communications to the lawyer, and (b) the AI program’s own commentary on what the lawyer has said, especially when, as in Heppner, the AI has been consulted without the lawyer’s permission and direction.
Furthermore, the client who shares attorney-client communications with AI, making them subject to disclosure, creates an additional headache requiring the collection of the files so they can be disclosed, a task that is not necessarily easy, straightforward or inexpensive, given the unintuitive way in which AI communications are often stored on a computer.[7] And of course, the typical lawyer’s compulsive motion practice to try to stop the other side from reaching into his client’s AI communications may run up fees for the client, waste the lawyer’s time, and delay resolution of the case.
The forgoing demonstrates that clients discuss their cases with AI at their own risk, and the lawyer who lacks (or loses) control over the client may be placed in the embarrassing position of having to force the client to turn over factual analysis, legal advice, and settlement positions to the adversary.
What does this mean for the lawyer’s obligations? Let’s turn to the language of the Rules of Professional Conduct: “A lawyer shall make reasonable efforts to prevent the inadvertent or unauthorized disclosure or use of, or unauthorized access to, information protected by Rules 1.6, 1.9(c), or 1.18(b).”[8] Lawyers are required “to make reasonable efforts to safeguard confidential information against unauthorized access by third parties and against inadvertent or unauthorized disclosure by the lawyer or other persons who are participating in the representation of the client or who are otherwise subject to the lawyer’s supervision.”[9] Surely a client participates in his own representation. A lawyer therefore has a duty to take reasonable measures to prevent his or her client from accidentally breaking the attorney-client privilege; if the privilege will be broken, the client should do it intentionally. “A lawyer shall explain a matter to the extent reasonably necessary to permit the client to make informed decisions regarding the representation.”[10]
The New York City Bar Association has recently issued a formal opinion specifically recommending that “[i]f an attorney knows that a client is recording a call with an AI tool, the lawyer should advise the client of the disadvantages of doing so.”[11] While there is no strict requirement that attorneys must discuss the risks and drawbacks of clients’ using AI to record attorney-client communications, it is considered best practice to raise these issues at the outset of the relationship — ideally when the engagement agreement is being formed. This is especially important if the attorney believes that the client may be inclined to record conversations or input communications into AI systems. Setting clear expectations from the beginning can often prevent future problems.[12] The lawyer may choose to include in an engagement letter an outright ban on the client’s recording the lawyer (with AI or otherwise), “provided the letter of engagement or retainer agreement: (a) does not mislead the client regarding circumstances under which the lawyer may seek to withdraw from the representation; and (b) does not serve as irrevocable advance consent by the client to withdrawal by the lawyer if the client violates the prohibition.”[13] Alternatively, the engagement letter may allow the client to use AI to record and make summaries of the lawyer’s communications only under certain conditions, for example, after the lawyer and client have investigated AI software and identified a particular configuration of AI software that will not break privilege when spoken to, like a purely internal private room offered by the AI software with custom-tailored protection for secrets spoken in it.
The following is an example of sample language that can be added to an engagement letter:
It is agreed that unless consent is obtained in advance, neither you nor the firm will use AI tools or other means to record, summarize or create transcripts of audio and video conversations with each other, including, without limitation, telephone, Zoom and Microsoft Teams calls. It is further agreed that neither you nor the firm will rely on any factual representation or purported legal advice that may be set forth in any such AI-generated recordings, transcripts or summaries unless they have been reviewed by both you and the firm to confirm the accuracy of the statements contained therein. Finally, to the extent that any such AI-generated recordings, transcripts or summaries are generated, both the firm and client must take special care to maintain confidentiality of the AI-generated document to avoid any loss of evidentiary privilege accorded to attorney-client communications or attorney work product and to avoid disclosure of client confidences.
Lastly, and perhaps the most problematic, this is a rapidly evolving area of practice with new decisions and articles written almost every day, so stay tuned for more.[14] If you or your clients are going to use AI, then your professional duty of competence[15] requires that you familiarize yourself to the appropriate extent with generative AI: how it works, how to configure it in such a manner that speaking to it preserves privilege and confidentiality and how to disclose it if need be. Computer technology has advanced to the point where it may no longer be true to stereotype the younger generation as computer savvy; a great percentage must be playing catch-up. But by putting in the effort to become competent and not fall too far behind, by attending continuing legal education courses on the subject, you will be ready to deal proactively and ethically with the new and unpredictable problems brought on by clients’ experimentation with AI, and indeed, with your own experimentation. After all, the only thing worse than your client’s accidentally forfeiting attorney-client privilege by talking to AI, is your own doing so!
Sincerely,
The Forum, by:
Vincent J. Syracuse, VSyracuse@thsh.com
Alyssa C. Goldrich, AGoldrich@thsh.com
Jean-Claude Mazzola, jeanclaude@mazzolalindstrom.com
Adam Wiener, adam@mazzolalindstrom.com
QUESTION FOR THE NEXT FORUM
To the Forum:
We all know that when we draft documents if we discover scrivener’s errors or mistakes of fact that deviate from what was agreed by the parties, we are required to correct those. But I have what is arguably a more complicated question.
I represent one of two individuals who are looking to form a corporation or limited liability company. My client will provide the expertise and will manage the entity, while the other individual will be an investor without much operating connection to the business. Once the entity is formed, they expect to raise additional money by selling shares in a private placement.
All of the business is expected to be conducted in New York. Both of the founders are New York residents and it is likely that the majority of the subsequent investment money will also come from New York residents. Accordingly, one might think that the most logical place to form the entity is under New York law.
However, in the interest of giving my client the greatest degree of flexibility and protection against being sued over how he manages the entity as a director or manager, as the case may be, I would like to put in the governing documents a provision that waives the applicable fiduciary duties to the greatest extent possible. I can achieve more of what I want in that regard under Delaware law than under New York law, and I can achieve even more under Texas law.
Accordingly, I would like to suggest to the other founder’s lawyer that we form the entity in either Delaware or Texas. I don’t want to disclose the aforementioned reason for why I want to do this, but rather if asked why, I can answer that Delaware has the most established jurisprudence for business entities or that Texas is trying very hard to accommodate companies and attract them to incorporate there and use its courts. While that might not be the “whole truth,” I believe that both of those explanations would be true and of non-trivial significance.
In the same vein, I would rather use an LLC if I determine that it is more flexible in this connection than a corporation. My preferred explanation would be that using an LLC plainly will allow the business to be run as a pass-through for tax purposes and if and when it becomes preferable to use an S corporation or even a regular C corporation, an LLC is able to elect that tax treatment. Again, it’s completely truthful and substantive but not a disclosure of what really matters to me and my client.
So my first question is whether there is anything in the Rules of Professional Conduct or otherwise that precludes me from doing what I want without making a full disclosure to the other founder’s lawyer that the fiduciary protection will be significantly less than with a New York corporation.
But wait, it’s actually more complicated than that. The other founder wants to have a right to purchase whatever New York real property the entity will acquire. My client does not like this element of the transaction but decided that he cannot fight it. You can imagine my barely concealed glee when I received the document drafted by the other founder’s lawyer providing for a 25-year option. Boom. I think that that option could well be void under the Rule Against Perpetuities.
So my second question concerns the misimpression that the other founder’s lawyer has created for himself without any fault, action or influence on my part: Am I required to inform him that this provision that he has proposed, which I know is so important to his client, is or may be a legal nullity?
Sincerely,
Jen Nanigans
Vincent J. Syracuse is a founding partner of Tannenbaum Helpern’s litigation and dispute resolution practice and has 50 years of experience in litigation. He received NYSBA’s Sanford D. Levy Professional Ethics Award and has chaired NYSBA’s program on ethics and civility for over 20 years. He co-chairs the Ethics Committee of the Commercial and Federal Litigation Section. He has been a co-author of the Attorney Professionalism Forum since 2012, which was published in a collection in 2021.
Alyssa C. Goldrich is an associate in Tannenbaum Helpern’s litigation and dispute resolution practice. She focuses her practice on complex commercial litigation matters in state and federal courts, with an emphasis on business torts, breach of contracts, intellectual property disputes, and real estate and construction litigation.
Jean-Claude Mazzola is founding partner of Mazzola Lindstrom LLP with over 25 years of experience as a commercial litigator. He is chair of NYSBA’s Committee on Attorney Professionalism.
Adam Wiener is an associate attorney at Mazzola Lindstrom LLP, where he focuses on contract law, bankruptcy, real estate finance, defamation, and constitutional law.
Endnotes:
[1] See, e.g., Robert Jablon and Jonathan Lloyd, Charles Manson Follower Leslie Van Houten Released From California Prison, NBC Los Angeles (July 11, 2023), https://www.nbclosangeles.com/news/california-news/charles-manson-follower-leslie-van-houten-released-from-california-prison/3185823. (“‘She’s been in prison for 53 years. She just needs to learn how to use an ATM machine, let alone a cell phone, let alone a computer,’ her attorney said.”).
[2] See The Age of AI and Our Human Future (2021), a book co-written by Henry Kissinger, Eric Schmidt (former CEO of Google) and Daniel Huttenlocher (dean of MIT’s College of Computing).
[3] See N.Y.C. Bar Ass’n Comm. on Pro. Ethics, Formal Op. 2003-02 (2003), https://www.nycbar.org/reports/formal-opinion-2003-02-undisclosed-taping-of-conversations-by-lawyers/.
[4] See, e.g., Xavier Rodriguez, Judging AI: How U.S. Judges Can Harness Generative AI Without Compromising Justice, 109 Judicature 10, 10–17 (2025), https://judicature.duke.edu/articles/judging-ai-generative-ai-courts. The article is written by a U.S. district judge who uses AI to manage his cases.
[5] No. 25 Cr. 503 (JSR) (S.D.N.Y., Feb. 17, 2026).
[6] No. 2:24-cv-12333 (E.D. Mich., Feb. 10, 2026).
[7] The mechanics of this are beyond the scope of this article, but see Staci D. Kaliner, Martin Tully & John P. Collins, Aligning Microsoft Tools with NYC Bar AI Recording Guidance, Redgrave LLP (Mar. 2026), https://www.redgravellp.com/publication/aligning-microsoft-tools-with-nyc-bar-ai-recording-guidance.
[8] N.Y. RPC 1.6(c).
[9] Id., cmt. 16.
[10] N.Y. RPC 1.4(b).
[11] N.Y.C. Bar Ass’n Comm. on Pro. Ethics, Formal Op. 2025-6. https://www.nycbar.org/reports/formal-opinion-2025-6-ethical-issues-affecting-use-of-ai-to-record-transcribe-and-summarize-conversations-with-clients/.
[12] Id.
[13] NYSBA Ethics Opinion 1270: Engagement Letters; Recording Attorney-Client Communications; Withdrawing From Employment; Advance Consent to Withdrawal. https://nysba.org/ethics-opinion-1270-engagement-letters-recording-attorney-client-communications-withdrawing-from-employment-advance-consent-to-withdrawal/.
[14] See e.g., Vincent J. Syracuse and Alyssa C. Goldrich. Attorney Professionalism Forum: Using AI in Your Practice? Proceed With Caution, NYSBA Journal, (September/October 2023), 54. https://nysba.org/using-ai-in-your-practice-proceed-with-caution/
[15] N.Y. RPC 1.1.
