Answers · Your case & AI
Before You Tell an AI Chatbot About Your Accident Case, Read This
A federal court has now ruled that conversations with AI chatbots are not privileged, not confidential, and can be obtained by the other side and used against you. If you are in a lawsuit, expecting one, or dealing with an insurance claim, what you type into ChatGPT, Claude, Gemini, or any similar tool may become evidence.
It is completely understandable why people do it. You have been hurt. You have questions at midnight when no lawyer is answering the phone. The chatbot is free, it answers instantly, and typing into that little box feels exactly like a private conversation.
It is not one. And a federal judge in New York just made that official.
Are AI chatbot conversations protected by attorney-client privilege?
No. A federal court has held that chatbot conversations carry neither attorney-client privilege nor work product protection. The other side can obtain them and use them against you.
In February 2026, Judge Jed Rakoff of the United States District Court for the Southern District of New York decided United States v. Heppner, what appears to be the first federal ruling in the country directly addressing whether conversations with an AI chatbot are protected by attorney-client privilege or the work product doctrine.
The answer was no. On both.
The facts should make anyone with a pending legal matter sit up. The defendant, a financial executive under federal investigation, had already hired an experienced law firm to defend him. He then used a consumer AI chatbot to research legal questions about the investigation, and he fed information he had learned from his own defense lawyers into the tool, generating dozens of documents of prompts and responses. Prosecutors sought those documents. The court let them have them.
The reasoning matters for everyone, not just criminal defendants:
An AI tool is not a lawyer. No attorney-client relationship can exist between a person and a chatbot. The privilege that protects what you tell your lawyer simply never attaches to what you tell a machine, no matter how legal the conversation feels.
You have no reasonable expectation of confidentiality. The platforms’ own terms permit collection and use of what you type. A conversation you have knowingly shared with a technology company is not a confidential one in the eyes of the law.
Worse: you can lose protection you already had. This is the part people miss. When the defendant typed his lawyers’ advice into the chatbot, the court treated that as a voluntary disclosure to a third party that waived the privilege over that information. Telling the AI what your lawyer told you is legally similar to reading your lawyer’s advice aloud to a stranger. The protection your attorney’s advice carried can evaporate the moment you paste it into a prompt.
And a paid subscription changes none of this. The court’s reasoning applied to the consumer product regardless of whether the user paid for it.
Can the defense get my ChatGPT conversations in an injury case?
Yes, where they are relevant. Heppner was a criminal case. Its logic applies squarely to civil litigation, and courts have already ordered AI chat logs produced. A candid 2 a.m. conversation about whether your neck really hurts as much as you testified is exactly the document a defense lawyer wants.
Heppner was a criminal case, but its logic applies squarely to civil litigation, including the auto negligence and insurance cases I handle. Consider what people routinely type into chatbots after a collision:
- Descriptions of how the crash happened, including details and doubts they have never told anyone
- Candid statements about their injuries: what genuinely hurts, what has improved, what they can and cannot do
- Questions that presume facts: “was it partly my fault if I was looking at my phone”
- What their lawyer told them, pasted in for a “second opinion”
- Settlement numbers, negotiation strategy, what they would really accept
- Frustrations, exaggerations, and thinking-out-loud that reads very differently on paper
Every line of that is potentially discoverable. In litigation, the defense can request your relevant documents and communications, and courts have already ordered the production of AI chat logs when they are relevant to the case. A candid 2 a.m. conversation with a chatbot about whether your neck really hurts as much as you claimed at your deposition is exactly the kind of document a defense lawyer dreams about.
And understand what discovery obligations mean once a case is underway: those chat logs may be evidence you are required to preserve. Deleting them after litigation is reasonably anticipated is not a solution: destroying relevant evidence is spoliation, and courts punish it severely, sometimes more severely than whatever the deleted material said.
Does incognito mode or a paid plan make a chatbot conversation private?
No. No privacy setting, incognito mode, or subscription tier creates a legal privilege. Privilege comes from the kind of relationship the law protects. Your relationship with a software product is not one of them.
The interface is designed to feel like a conversation. Conversations feel private. But legally, typing into a chatbot is closer to writing a letter and handing a copy to a technology company than it is to talking with a lawyer, a doctor, or a spouse.
Those three get privileges because the law has decided those relationships deserve protection. Your relationship with a software product enjoys no such protection, and no privacy setting, incognito mode, or subscription tier creates one. The Heppner court left open only the possibility that enterprise-grade tools operating under strict confidentiality agreements, used at a lawyer’s direction, might someday be treated differently. That is not the app on your phone.
Is it safe to use AI to research the law at all?
Yes, as long as you keep it general. Asking what Michigan’s no-fault system is, or what serious impairment of body function means, involves no facts about you. Nothing is at risk. The line is specifics. The moment the conversation turns to your crash, your injuries, or your lawyer’s advice, you are creating an unprotected written record about your case.
To be clear about what this article is not saying: there is nothing wrong with using AI to learn about the law in general. Asking a chatbot “what is Michigan’s no-fault system” or “what does serious impairment of body function mean” involves no facts about you. General legal education is what these tools are reasonably good at, and reading about the law puts nothing at risk.
The line is specifics. The moment the conversation turns to your crash, your injuries, your fault, your lawyer’s advice, or your settlement thinking, you are creating a written record about your case with no legal protection around it.
A simple test before you hit enter: would you be comfortable if the insurance company’s lawyer read this aloud at your deposition? Because that is now a realistic destination for it.
Where can I safely be completely honest about my case?
With your own lawyer. What you tell them in confidence, to get legal advice, carries one of the oldest and strongest shields in the law. That is what a chatbot cannot offer at any price: privilege.
What you tell your lawyer, in confidence, for the purpose of getting legal advice, is protected by one of the oldest and strongest shields in the law. That protection exists precisely so you can be completely honest, about the bad facts, the doubts, and the mistakes, and get real advice based on the truth. Your lawyer cannot help you around a problem they do not know exists, and the privilege is what makes telling them safe.
The midnight questions are real, and they deserve real answers. Write them down. Bring them to your lawyer. Every one of them will be answered inside a conversation the other side can never touch.
The machine answers instantly. The lawyer answers confidentially. In a lawsuit, only one of those matters.
Frequently asked questions
Are conversations with ChatGPT or Claude protected by attorney-client privilege?
No. In United States v. Heppner (S.D.N.Y. 2026), a federal court held that AI chatbot conversations are not protected by attorney-client privilege or the work product doctrine. An AI tool is not a lawyer, no attorney-client relationship can exist with a chatbot, and platform terms permitting collection of what you type defeat any reasonable expectation of confidentiality.
Can AI chat logs be subpoenaed or used in a lawsuit?
Yes. AI conversations relevant to a case are potentially discoverable like other documents and communications, and courts have ordered their production. Worse, typing your lawyer’s advice into a chatbot can waive the privilege that advice previously carried, as a voluntary disclosure to a third party.
Is it safe to ask AI general legal questions?
Generally yes. Asking about the law in general, such as what Michigan’s no-fault system is, involves no facts about you and puts nothing at risk. The line is specifics: the moment the conversation turns to your crash, your injuries, your fault, or your lawyer’s advice, you are creating an unprotected written record about your case.
Should I delete my AI chats if I’m in a lawsuit?
No. Once litigation is reasonably anticipated, relevant chat logs may be evidence you are required to preserve. Deleting them is spoliation, and courts punish destruction of evidence severely, sometimes more severely than whatever the deleted material said. Talk to a lawyer instead.
Talk to someone about your case
I am a trial attorney at Michigan Auto Law handling Michigan auto negligence and no-fault cases. If you have questions about a collision, an injury, or an insurance claim (including questions you have already asked a chatbot), bring them to a conversation the law actually protects. The consultation is free, and it is confidential in a way no AI conversation is.
This article discusses the law as of July 2026 and is general information, not legal advice. It does not create an attorney-client relationship. The law governing AI communications and discovery is developing rapidly, courts in other jurisdictions may reach different conclusions, and outcomes depend on the specific facts of each case. Attorney advertising. Gary Safir is a trial lawyer at Michigan Auto Law.