The short answer
If you represent yourself: usually not, but not always. As of October 2026, federal courts have held that a person representing themselves in a civil lawsuit can protect their prompts and responses in ChatGPT, Claude, Grok, and similar tools as "work product" when the chats were created to prepare their case. Using a free, public chatbot does not by itself give that protection away. But courts are also making clear that the protection is not automatic. Chats that were not made for the lawsuit, the facts you typed in, and the name of the AI tool you used can all be discoverable.
If you have a lawyer: it depends. The leading case, United States v. Heppner (S.D.N.Y. 2026), held that a criminal defendant who used consumer Claude on his own, without his lawyer's direction, got neither attorney-client privilege nor work-product protection. Other courts read the work-product rule more broadly, but none has held that a represented person's independent chatbot use is automatically protected. AI use at your lawyer's direction stands on much firmer ground.
Either way, the facts you typed in are not protected, and pasting your lawyer's advice into a public chatbot can waive the attorney-client privilege for that advice.
What "work product" means, and why it matters for AI users
Work product is the legal rule that keeps the other side from seeing how you prepare your case. In federal court it comes from Federal Rule of Civil Procedure 26(b)(3). That rule protects documents "prepared in anticipation of litigation or for trial by or for another party or its representative."
The rule says "party," not just "attorney." That matters because millions of people now use ChatGPT, Claude, Grok, Gemini, and similar tools to understand their rights, organize facts, and draft filings without a lawyer. When the other side asks for those chats in discovery, work product is usually the only shield available, and how it applies depends on whether you have a lawyer.
Attorney-client privilege does not help here. A chatbot is not a lawyer, and one federal court has already said so plainly: United States v. Heppner, 820 F. Supp. 3d 292, 296 (S.D.N.Y. 2026).
What courts decided in 2026
Six federal decisions and one New York state decision frame the issue. The trend is that a chatbot is treated as a tool, not a third party, but each chat must still qualify as litigation preparation.
Warner v. Gilbarco, Inc. (E.D. Mich., Feb. 10, 2026)
A self-represented employee's ChatGPT materials were protected work product, and using ChatGPT did not waive that protection. The court said AI tools "are tools, not persons." Citation: 820 F. Supp. 3d 629.
United States v. Heppner (S.D.N.Y., Feb. 17, 2026)
A criminal defendant who had a lawyer, but used consumer Claude on his own, got neither attorney-client privilege nor work-product protection for his AI chats. Citation: 820 F. Supp. 3d 292.
Morgan v. V2X, Inc. (D. Colo., Mar. 30, 2026)
A self-represented employee's AI use was protected work product, but he had to disclose which AI tool he used. The court also barred uploading the other side's confidential discovery into free AI tools. Citation: 2026 WL 864223.
Assini v. Hayward (N.Y. Sup. Ct., June 4, 2026)
Subpoenas to an AI company seeking a self-represented defendant's chats were quashed on work-product grounds. Citation: 2026 N.Y. Slip Op. 26086.
Villanueva v. Las Vegas Metropolitan Police Department (D. Nev., Sept. 29, 2026)
An expert witness's ChatGPT prompts were protected, but the AI-generated case summaries he relied on had to be produced. Using ChatGPT did not waive protection. Citation: 2026 WL 2942348.
Fontainebleau Florida Hotel, LLC v. Botach (S.D. Fla., Oct. 21, 2026)
A party's AI chats get no blanket protection; each chat must have been made because of the litigation. Chats used to create content about the other side were discoverable, and narrowed subpoenas to AI providers were allowed. Citation: 2026 WL 3151630.
Fischman v. Epic Systems Corp. (N.D. Tex., Oct. 23, 2026)
A self-represented plaintiff was barred from putting the other side's confidential documents into AI tools that train on user inputs. Citation: 2026 WL 3182476.
No federal appeals court has decided the question yet. Neither has any court in Illinois.
Does using free ChatGPT, Claude, or Grok waive protection?
In most federal courts, no. Work-product protection is waived only when material is disclosed to an adversary, or in a way that makes it substantially more likely an adversary will get it. Typing into a chatbot doesn't meet that test, courts reason, because the AI company is not your opponent and is unlikely to hand your chats to them without legal process.
The Warner court put it simply: ChatGPT and other generative AI programs "are tools, not persons, even if they may have administrators somewhere in the background." It added that treating chatbot use as waiver "would nullify work-product protection in nearly every modern drafting environment."
The free-version problem is real, though. OpenAI's own Data Controls page explains that conversations can be used to train its models unless you turn off "Improve the model for everyone." The Heppner court relied on a chatbot's privacy policy to find "no reasonable expectation of confidentiality" for attorney-client privilege purposes. Most civil courts have not applied that reasoning to work product, but the argument is out there, and a court could take it more seriously if your chats included someone else's confidential documents.
The same goes for other consumer chatbots. Anthropic (Claude) and xAI (Grok) each publish their own privacy policies with their own data-use and retention terms, and Heppner turned in part on the wording of Claude's policy. Before you rely on any free tool for your case, read its privacy settings and turn off model training if the option exists.
If you have a lawyer: that is different
Most of the decisions protecting AI chats involved people representing themselves. Courts reason that a self-represented person is "simultaneously the party and the advocate," so their own thinking about the case is the lawyer's thinking. When you have a lawyer, two separate questions arise.
Attorney-client privilege. Your conversations with your lawyer are privileged. Your conversations with a chatbot are not, because a chatbot is not a lawyer. In Heppner, the defendant typed information he had learned from his lawyers into Claude. The court held he waived privilege for that information "by sharing [it] with Claude and Anthropic, just as if he had shared it with any other third party." Do not paste your lawyer's emails, advice, or strategy into a public chatbot.
Work product. Heppner also held that the defendant's AI documents were not work product because they were not prepared "by or at the behest of counsel" and did not reflect his lawyer's strategy. The court suggested the result could be different if the lawyer had directed the AI use. Other courts read the rule more broadly. In Fontainebleau Fla. Hotel, LLC v. Botach (S.D. Fla. Oct. 21, 2026), a represented defendant's AI chats were not categorically unprotected, because the federal rule covers materials prepared "by or for another party." But that court still required him to show that each chat was made because of the lawsuit, and it allowed discovery of chats he used to create content about the other side. A Texas business court likewise protected a company principal's ChatGPT chats under the Texas rule.
The practical rule for represented people: use AI for your case only as your lawyer directs, preferably through tools your lawyer chooses that keep inputs confidential and do not train on them.
What is not protected
Work product protects your thinking about the case. It does not protect everything you ever typed into a chatbot. Based on the 2026 decisions, expect the other side to be able to get the following:
The facts. You can be asked at a deposition or in written questions what happened, even if you once described it to ChatGPT, Claude, or Grok. Facts are never work product.
Which AI tool you used. Morgan ordered a self-represented plaintiff to name the platforms he used with confidential information.
What documents you uploaded. A Texas court ordered a party to list, document by document, every discovery record it put into ChatGPT.
Chats that were not made for the lawsuit. Asking ChatGPT for advice at work, venting about a boss, or drafting an email before any dispute existed are not litigation preparation. Fontainebleau also held that chats used to create posts or statements about the other side are not work product.
AI output you relied on as evidence or shared with others. Quoting a chat in a filing, or sending it to a friend, an agency, or the other side, can waive protection for that material.
If you refuse to produce anything, courts increasingly require a privilege log. That is a list describing each withheld chat, without revealing its contents, so a judge can decide whether it qualifies.
What about Illinois courts?
No Illinois state court has decided whether a litigant's AI chats are discoverable. Illinois appellate decisions on AI so far deal with fake citations, not discovery. Examples are In re Baby Boy, 2025 IL App (4th) 241427, and Scott v. Illinois Human Rights Commission, 2026 IL App (1st) 251462.
Illinois work-product law is narrower than the federal rule. Illinois Supreme Court Rule 201(b)(2) protects only material that discloses "the theories, mental impressions, or litigation plans of the party's attorney." Read literally, that language protects a lawyer's thinking, not a self-represented person's. No Illinois court has yet said whether someone representing themselves counts as their own attorney under that rule. If your case is in Illinois state court, assume your AI chats have less protection than they would in federal court.
The Illinois Supreme Court's Policy on Artificial Intelligence (effective January 1, 2025) allows litigants to use AI and says disclosure of AI use "should not be required in a pleading." It also says "self-represented litigants are accountable for their final work product" and must "thoroughly review AI-generated content" before filing. The policy addresses filings, not whether your chat history can be discovered.
Cases in federal court in Illinois, such as federal employment-discrimination suits, follow the federal rule described above. The Seventh Circuit applies the same "substantially increases the opportunity for adversaries" waiver test the other courts use: Appleton Papers, Inc. v. EPA, 702 F.3d 1018, 1025 (7th Cir. 2012).
How to use AI in your own legal dispute without losing protection
These steps follow what courts have actually rewarded or punished in 2026 decisions.
If you represent yourself:
Keep case work in separate chats. Start a new conversation for each litigation task, and label it (for example, "Drafting my response to the motion"). That makes it easier to show a chat was made because of the lawsuit.
Don't mix the lawsuit with everyday use. Venting about your employer, asking for workplace advice, or drafting ordinary emails in the same thread weakens any work-product claim.
Never upload the other side's confidential documents. Courts are entering orders that bar putting confidential discovery into ChatGPT, Claude, Grok, or similar tools that may train on inputs. Violating a protective order can lead to sanctions.
Turn off model training where the tool allows it, and don't delete chats once a lawsuit is likely. Deleting history after you expect litigation can look like destroying evidence.
Don't share chats with friends, family, or agencies. Sharing can waive protection.
Check every citation yourself. Courts in Illinois and elsewhere are sanctioning self-represented parties for AI-invented cases, and "ChatGPT told me" can put your whole process at issue.
If you withhold chats in discovery, list them on a privilege log. Refusing to respond at all is not an option.
If you have a lawyer:
Ask your lawyer before you use any AI tool on your case. Use AI for case work only when your lawyer directs it.
Never paste your lawyer's emails, advice, or strategy into a public chatbot. That is the scenario Heppner treated as a waiver of attorney-client privilege.
Let your lawyer choose the tool. Business and enterprise versions that contractually bar training and disclosure are far easier to defend than free consumer versions.
Tell your lawyer what you have already done. If you used a chatbot before hiring counsel, your lawyer needs to know which tool, what you typed, and whether the history still exists.
Frequently asked questions
Can my ChatGPT, Claude, or Grok chats be used against me in a lawsuit?
Some can. Chats you created to prepare your case are often protected as work product, especially if you represent yourself. Chats made for other reasons, the facts you typed in, and the name of the tool you used can be discoverable.
Are AI chats protected by attorney-client privilege?
No. A chatbot is not a lawyer, so conversations with it are not attorney-client privileged. Sharing your lawyer's advice with a chatbot can also waive privilege over that advice (United States v. Heppner, S.D.N.Y. 2026).
Does using the free version of ChatGPT waive work-product protection?
Not by itself, according to most federal courts that have ruled. Courts treat the chatbot as a tool, not an adversary. But free versions may use your inputs for training, and courts may scrutinize that, especially if you uploaded someone else's confidential documents.
Is the answer different if I have a lawyer?
Yes. If you have a lawyer and use a chatbot on your own, at least one federal court has held your chats are not protected at all. AI use directed by your lawyer is more likely to be protected.
Can the other side subpoena OpenAI, Anthropic, or xAI for my chats?
They can try. A New York court quashed such subpoenas in 2026, while a Florida federal court allowed narrowed subpoenas to AI providers. Federal law (the Stored Communications Act) generally limits providers from disclosing message content in civil cases. So discovery usually goes through you.
Do I have to tell the court I used AI?
It depends on the court. The Illinois Supreme Court's policy says disclosure "should not be required in a pleading," but some individual federal judges require AI disclosures. In every court, you are responsible for checking what you file.
What should I do if I already used a chatbot for my case?
Don't delete the history. Stop uploading confidential documents, and talk to a lawyer about what you have and how to protect it.
About the author
Jon Phillips is an Illinois litigation and intellectual property attorney with Brown, Hay & Stephens, LLP in Peoria. He has argued before the U.S. Court of Appeals for the Federal Circuit and litigates in federal and state courts across Illinois. To discuss how AI use affects your case, contact Jon at (309) 643-6518 or jphillips@bhslaw.com.
This article is general information current as of October 24, 2026, not legal advice. Reading it does not create an attorney-client relationship. The law on AI and discovery is changing quickly; talk to a lawyer about your specific situation.