ARE YOUR AI SEARCHES GOING TO BE USED AS EVIDENCE AGAINST YOU IN YOUR CASE?

Did you use an AI engine to find out about your case before you hired an attorney?  If so, you may be at risk of having those AI searches revealed as part of the discovery in your case.  Discovery is the process of obtaining information about your case.  Discovery comes in four flavors: interrogatories, request for production of documents, requests for admissions and depositions.  Each of these provide a different way of obtaining information. Interrogatories are questions. Requests for production asks you to produce documents that the other side wants to see if you have. Requests for admissions ask you if specific statements are true, or if you admit that a document is a true and correct copy of a document at issue in the litigation. Depositions are where the attorney gets to ask questions orally of a witness.  The parties to the case are subject to all forms of discovery.  Third parties are subject only to requests for production of documents and depositions. In each of these events, your lawyer can object to the other side asking for any private communications you and the attorney have or have had, based on the attorney client privilege, so long as the conversation was private and kept private between the attorney and the client. If the opposing party asks how you reached the conclusions or theories that you reached, your attorney can object based on work product doctrine - i.e. that it invades the attorneys' thought process. 

 

Unfortunately, now that people are putting into artificial intelligence programs, their questions about their case - so that they can figure out how to help themselves or perhaps understand what type of attorney might best help them, there is a significant question about whether or not opposing counsel can ask the party, or the artificial intelligence engine company about any such searches.  What? How? Well, let's assume that you become aware you might be, or are, being sued.  You go to your computer and ask Anthropic's Claude a series of questions based upon what you know about your case.  Perhaps something like "Is it a problem if I knew X before I sold the business?"  or "Is it a problem that I knew X, did not tell any of my partners or business associates, and then bought the business for myself? " or What would a party need to  to establish intent? "  These questions lay out the areas that you are most concerned about.  These questions can, depending on their specificity, give the opposing party the very elements of scienter - knowledge - that the opposing party needs to prove their case. 

 

These issues have started to crop up in cases.  In United States v. Heppner, 2026 U.S.Dist. Lexis 32697 (SDNY 2/17/2026) Judge Jed S. Rakoff required the defense to turn over the pre-hiring searches of Anthropic's Claude to him so he could analyze this very question.  Because the Defendant had not yet hired counsel when he used Claude, and in part, because of the high standard for turnover in criminal cases, Judge Rakoff turned over the searches to the prosecution.  Judge Rakoff reasoned that no because the work product doctrine is to protect lawyer's mental processes, and that no attorney client relationship existed at the time of the searches, and that Mr. Heppner had conducted the searches himself without counsel, that the Claude searches were not protected. 

 

Heppner was decided on February 17, 2026.  On March 30, 2026, Judge Maritza Dominquez Braswell, of the United States District Court for Colorado, reached the opposite conclusion, finding that artificial intelligence searches were not subject to turnover through discovery.  The critical difference in Judge Braswell's decision was that the case was a civil matter, being decided in the context of an employment case under the Federal Rules of Civil Procedure rather than the Federal Rules of Criminal Procedure.  The FRCP have a broader scope of protections related to mental impressions and litigation preparation materials under FRCP 26(b)(3).  Further, the Plaintiff was representing himself, and Judge Braswell interpreted the attorney client privilege to apply to the self-represented individual. Lastly, Judge Braswell concluded that the gap between the Heppner defendants searches and the hiring of counsel presented a significantly different landscape to apply the rule to, than the case before her, in which the self-represented litigant was handling the matter. 

 

Where does that leave litigants and lawyers?  At this time there is no clear standard, certainly not in the 9^th Circuit or California.  However, this issues is likely to crop up in the future if not in cases that are on-going at the moment. Attorneys should certainly know if their clients have been using AI before they came to the attorney, and should caution their clients about the use of AI and potential risk that such AI might subject their concerns and searches to review by the opposing counsel.  Arguably attorneys should start including in their discovery requests questions about their opponents use of AI, in the hopes of obtaining information about the motives and thinking of the opposing party.