If an AI chatbot told you what was wrong with you, told you what to do about it, and you were hurt — that conversation may matter more than you think. It may be the start of a ChatGPT lawsuit.
Key takeaways
- California licenses the practice of medicine. The law asks what an actor did, not what the actor called itself.
- A chatbot that names a condition, rules one out, or directs a prescription drug has done something California reserves to licensed clinicians.
- A line reading “I’m not a doctor” addresses one part of the licensing statute. It does not answer whether a diagnosis or a treatment direction actually happened.
- Your conversation history is the evidence. Export and preserve it before anything else.
- Review costs nothing, and the firm is paid only out of a recovery.
What these cases are about
Roughly 230 million people bring health questions to ChatGPT every week, by OpenAI’s own count. Most of those conversations do no harm. Some end somewhere else entirely — a missed pulmonary embolism, a stopped psychiatric medication, an emergency room visit that happened days later than it should have. Those are the conversations that become a ChatGPT lawsuit.
The question these cases raise sounds simple and is not: when a general-purpose chatbot stops describing a disease and starts telling you what you have and what you should do about it, has it crossed into something the law regulates?
California has answered a version of that question for over a century. Business and Professions Code sections 2051 and 2052 make it unlawful for an unlicensed person to diagnose, treat, or prescribe for any physical or mental condition. The statute reaches conduct. California courts have described the practice of medicine as “very broad,” covering any activity customarily performed by a licensed practitioner — including, expressly, giving advice and counsel on the taking of medications.
That body of law was not written with software in mind. It was written to reach whoever was performing the conduct.
The patterns that matter
Not every health conversation with an AI raises a legal question, and not every bad answer supports a ChatGPT lawsuit. In the records we review, a handful of patterns recur — and these are the ones worth a closer look.
It named a condition. The chatbot told you what you had. Under California’s older statutory definition of diagnosis, still quoted by the courts, it makes no difference whether the undertaking was paid or free.
It ruled a condition out. “This isn’t a clot.” “You’re not having a heart attack.” “You’re not manic.” A negative finding is a finding. The definition reaches any undertaking to establish whether or not a person is suffering from a disorder — whether or not is the operative phrase, and courts have applied it to exactly that situation.
It directed a prescription medication. Start one, stop one, skip a dose, double a dose, taper off, flush them. No therapist in California may do any of this. The practice of psychology statute says so in plain terms: psychologists may not prescribe. Medication direction sits with physicians, and with physician assistants, nurse practitioners, and pharmacists operating under defined agreements — nobody else.
It told you to wait. Some conditions are measured in hours. Pulmonary embolism, stroke, heart attack, sepsis, appendicitis, diabetic ketoacidosis, testicular or ovarian torsion. Advice that reframes those symptoms as minor, or sets a threshold you should hit before seeking care, is advice with a clock running against it.
It displaced your doctor. You told it you had no physician, could not afford one, or were asking it instead of going in. Or it coached you on what to tell a clinician, or suggested your doctor had it wrong. California’s relationship cases turn on reliance — on whether the speaker knew the listener would act on what was said.
The disclaimer question
Nearly every one of these conversations contains a caveat somewhere. “I’m not a doctor.” “Consult a healthcare professional.”
Those lines matter as evidence, and not in the way you might assume.
California’s medical practice statute has two separate prongs: one reaching those who hold themselves out as practicing, and a different one reaching those who actually diagnose, treat, or prescribe. A disclaimer speaks to the first. It says nothing about whether the second happened. A California court addressing this point held that a violation “is not absolved by a concurrent statement that the violator is not a doctor.” (People v. Cantor (1961) 198 Cal.App.2d Supp. 843, 848.)
The Legislature has also written the only disclosure safe harbor that exists in this area — and it demands a signed, written statement given before services, retained for three years. Even then it does not cover recommending that someone discontinue a prescribed drug, and it does not cover diagnosing or treating under conditions that create a risk of serious harm. A conversational aside is not a near-miss of that standard. It is the thing the Legislature considered and declined to accept.
There is a further wrinkle worth knowing. In some records the caveats thin out as the conversation goes on, and eventually stop appearing. OpenAI has publicly acknowledged that its safeguards “can sometimes be less reliable in long interactions” because parts of the model’s safety training may degrade. Where that pattern shows up in a transcript, it tends to be one of the more telling things in the file.
What has already been filed (updated September 2026)
These are not hypothetical claims. Several ChatGPT lawsuits are on file in California and are matters of public record. The ChatGPT lawsuits filed so far fall into recognizable groups: suicide and self-harm, delusion and mental health crisis, third-party harm, and a newer group concerning medical advice and treatment direction. This section covers the medical group.
The California coordinated proceeding
ChatGPT lawsuit filings against OpenAI in California state court are coordinated before the San Francisco Superior Court as Judicial Council Coordination Proceeding No. 5431, In re: ChatGPT Product Liability Cases, assigned to Judge Ethan P. Schulman. Coordination gathers related cases before one judge for pretrial purposes. It is not a class action; each plaintiff keeps their own case. The Schenk Law Firm Partner, Benjamin S. Schenk, was appointed by Judge Schulman to serve on the Plaintiff’s Steering Committee, a leadership role in that proceeding.
Winters v. OpenAI: delayed emergency care
Filed: July 2026, San Francisco Superior Court. Defendants: OpenAI entities; Sam Altman individually.
A Florida pastor alleges he consulted ChatGPT-4o about recurring dizziness and blood pressure instability, and later about groin pain. The complaint alleges the chatbot treated the symptoms as relatively minor, told him the pain was “very likely another minor piece of the long story,” advised him to remain “recliner-bound,” and set a threshold of roughly eight to ten further episodes before the condition would count as serious. He suffered a massive pulmonary embolism — clots in both lungs.
Claims include: negligence, the unauthorized practice of medicine, strict liability, violation of California’s Unfair Competition Law, and negligent undertaking against Altman. The New York Times first reported it as a ChatGPT medical advice lawsuit. The complaint also alleges that early answers carried referrals to a healthcare provider and that the caveats thinned as the conversation lengthened.
Turner-Scott v. OpenAI: medication direction
Filed: May 12, 2026, San Francisco Superior Court. A wrongful-death action brought by the family of a 19-year-old who, the complaint alleges, was directed by ChatGPT to take Xanax to manage nausea during kratom withdrawal, and died. Pleads negligence grounded in California’s professional licensing statutes, with the same unlicensed-practice framing, and seeks an order pausing ChatGPT Health.
Pennsylvania v. Character Technologies: the holding-out case
Filed: May 1, 2026, Commonwealth Court of Pennsylvania, by the State Board of Medicine. Alleges a Character.AI chatbot presented itself as a licensed clinician — describing itself as a “Doctor of psychiatry,” asserting that assessing for medication was “within my remit as a Doctor,” and supplying a fabricated license number. Misrepresenting a license is not a close constitutional question, and it is what California’s section 4999.9 now reaches, with each use a separate violation. A December 2025 coalition letter from 42 attorneys general warned AI companies that chatbots have been linked to multiple deaths and may face liability under existing state law.
Legislatures have moved in the same direction as the OpenAI lawsuits
Nevada now bars AI systems programmed to provide services that would constitute professional mental or behavioral health care if a person provided them. Illinois bars internet-based AI from offering therapy unless a licensed professional conducts it. Utah, Maine, Rhode Island, Tennessee, and New York have enacted parallel measures. As of January 1, 2026, California’s section 4999.9 reaches anyone who develops or deploys an AI system using protected professional titles, with each use a separate violation. California’s SB 243 (Business and Professions Code sections 22601 to 22605) adds a companion-chatbot statute with a private right of action, and a pending New York bill would bar AI from giving medical or legal advice and provides that a displayed disclaimer is no defense.
What the courts have not decided yet
- No appellate ruling yet on whether licensing statutes reach a general-purpose AI product.
- The First Amendment defense is untested in this context. A June 2026 federal decision held counselor licensing unconstitutional as applied to talk therapy, while expressly preserving licensing for those who “prescribe medicine.”
- Retroactivity. California’s 2026 statutes do not reach earlier conduct. The practice of medicine statutes have been on the books far longer.
Summarized from public filings and published reporting. Allegations have not been proven. Nothing here predicts an outcome in any case.
What to preserve, starting now
Conversation history is the case. Before anything else:
- Export your account data. In ChatGPT, that lives under Settings → Data Controls → Export Data. You will receive an email with a download link. Full step-by-step instructions here.
- Do not delete the conversation, and do not delete the account.
- Screenshot the specific exchanges — the question you asked and the answer you got — with dates visible.
- Collect your medical records for the same window, including any visit that came after.
- Write down what you did because of the advice: a dose changed, an appointment skipped, a trip to the ER postponed.
- Note which product and mode you used — ChatGPT, ChatGPT Health, a custom GPT, Character.AI, a companion app, a persona with a clinical name.
Who should have a conversation reviewed
You do not need to know whether you have a case. That is the review’s job. It is worth a call if:
- A chatbot named a condition you had, or told you that you did not have one you turned out to have.
- It directed you to start, stop, change, or skip a prescription medication.
- It told you your symptoms did not require urgent care, and they did.
- You told it you had no doctor or could not afford one.
- You asked it to act as your therapist or physician, and it accepted the role.
- Someone in your family died, and an AI conversation sits anywhere in the timeline.
Why this firm
The Schenk Law Firm is a San Diego personal injury and mass tort firm that reviews AI medical advice claims nationwide. Attorney Benjamin S. Schenk was appointed by Hon. Ethan P. Schulman to serve as a member of the Plaintiffs’ Steering Committee, a leadership role in the California coordinated proceeding covering ChatGPT product liability cases. The Schenk Law Firm has recovered more than $25 billion on behalf of clients. Past results do not guarantee future outcomes.
We are experienced with emerging-technology injury claims — underage online gambling, social media addiction, and AI-related mental health litigation among them. Cases involving AI medical advice are new, the law governing them is developing, and no lawyer can promise an outcome in any of them. What we can tell you is whether a record shows the pattern a ChatGPT lawsuit requires.
Talk to us
Case review costs nothing and carries no obligation. We work on contingency: no fee unless there is a recovery.
The Schenk Law Firm, LLP 3180 University Avenue, Suite 260, San Diego, CA 92104 (858) 424-4444
If you are in crisis, call or text 988 to reach the Suicide and Crisis Lifeline. If you have symptoms that may be an emergency, call 911 or go to an emergency room.
Frequently asked questions
Can I sue ChatGPT for medical advice?
You would not sue the product; you would sue the company that built and deployed it. Complaints on file in California name OpenAI entities and, in some cases, Sam Altman individually. An AI chatbot lawsuit of this kind is brought against the developer and deployer, not against the software; the OpenAI lawsuits described above show what that looks like.
Does it matter that I used the free version?
For the licensing theory, no. California removed “for a fee” from its practice-of-psychology definition effective January 1, 2016, and the medical practice statute’s definition of diagnosis reaches undertakings “whether gratuitous or not.” Payment can still matter for other claims.
The chatbot told me to see a doctor. Does that end it?
Not by itself. A referral followed by substantive direction is a different record from a referral alone. Both halves matter.
What if I have a psychiatric history?
That history is not a reason to stay quiet. In these records it is frequently the reason the conversation went where it did.
How long do I have?
California personal injury deadlines are short and vary by claim and by who was harmed. Do not wait to find out which one applies to you.
Is an AI chatbot lawsuit different from a medical malpractice case?
Yes. Malpractice requires a licensed clinician and a doctor-patient relationship. These cases plead that a company let a product do what only a licensed clinician may do, which is why the licensing statutes, not the malpractice standard, set the measure of care.
