By now, most attorneys have certainly seen how AI can supplement one’s practice in any number of ways. But, what happens when AI is used against you?
One trend I've encountered with striking regularity, sometimes daily, is the surge in pro se filings that are unmistakably generated by AI tools such as ChatGPT, Claude, and similar platforms. Responding to these filings presents a genuine challenge: AI enables pro se litigants to produce a high volume of submissions in a compressed timeframe. More notably, the filings are often well-crafted and raise complex legal questions that warrant substantive, thorough responses.
I have spent considerable time wondering about how to guard against the rise of frivolous AI filings, including whether these filings need to be combatted at all. It appears that we are in a regulatory lag whereby state bar associations, legislatures, and even courts have not had the time to address the rise of these types of filings.
One recent case from the United States District Court for the Southern District of New York addressed whether AI chat records are entitled to protection via the attorney-client privilege or work product doctrine. See United States v. Heppner, 2026 U.S. Dist. LEXIS 32697, at *6 (S.D.N.Y. 2026). In answering that question in the negative with respect to the attorney-client privilege, the United States District Court was persuaded by the fact that (1) the AI platform in question (Claude) was not an attorney; (2) the platform indicated that the submissions were not confidential; and (3) the user inputs were not made at the direction of an attorney. See id. at 6-9. With respect to the work-product doctrine, because the party chatting with Claude was not acting at the direction of an attorney, the Court found the doctrine inapplicable. Id. at 13-14. (It is worth noting that the party chatting with Claude was represented by an attorney, but the attorney was not involved in the chat process).
While it does not appear as of the time of this writing that a Texas appellate court has weighed in on this question, I wonder how the Heppner ruling would compare with a Texas determination, when measured against Texas’ own privilege and work product frameworks.
On the privilege side, the result would likely be the same. Under Texas Rules of Evidence 503, a communication is generally attorney-client privileged when it is between a client and the attorney, or between or among their representatives. See Tex. R. Evid. 503(b)(1). And, each of the definitions of “client's representative” and “lawyer’s representative” necessarily require a “person.” See Tex. R. Evid. 503(a). Since AI is not a person, pro se chats with Claude to develop litigation strategy would likely not be privileged under Texas law for the same fundamental reason identified in Heppner: no attorney-client relationship exists.
The work product question, however, is where Texas law and the Heppner framework could potentially diverge. The distinction lies in how each defines work product. The Heppner court applied an attorney-centric concept of the doctrine—one designed to shelter the mental processes of an attorney or actions done at the direction of an attorney. Even though Heppner was represented, because he acted on his own, the court found no work product protection.
Texas Rule 192.5(a)(1) asks a different question. It protects material “prepared or mental impressions developed in anticipation of litigation or for trial by or for a party or a party's representatives, including the party’s attorneys, consultants, sureties, indemnitors, insurers, employees, or agents.” The “party” is independently listed. Further, even under Texas law, certain work product can still be discovered upon a showing that the requesting party has a substantial need for the materials and cannot obtain their substantial equivalent without undue hardship. See Tex. R. Civ. P. 192.5(b)(2). Contrast this with “core work product,” which does involve an attorney and is not discoverable.
While it may be some time before we see a Heppner-like decision in Texas, it will be interesting to see how a Texas court construes privilege issues and AI, especially in the context of pro-se litigation and the seemingly broader work product doctrine.