AI Psychosis

AI psychosis is a plain-language term, not a medical diagnosis. People use it when someone's long conversations with a chatbot seem to feed false beliefs, paranoia or a break from reality. Lawsuits filed in California since 2025 claim that OpenAI designed ChatGPT in ways that made this more likely, then failed to warn users. OpenAI disputes the claims, and no court has found it liable. An AI psychosis lawyer can look at the chat history and medical records and explain whether an AI psychosis lawsuit may fit what happened.

What is AI psychosis?

Psychosis is a set of symptoms in which a person loses some contact with reality. Thoughts and perceptions are disrupted, and it can be hard to tell what is real1. Common symptoms are delusions, which are false beliefs, and hallucinations, which means seeing or hearing things others do not1.

The lawsuits describe something narrower. They say a chatbot mirrored and agreed with a user's ideas, even when those ideas were false or dangerous, until the user believed them more strongly. The complaints call the result "AI-related delusional disorder"4. That phrase comes from the lawsuits. It is not a recognized diagnosis.

OpenAI has published its own estimate. In October 2025 it said about 0.07% of users active in a given week showed possible signs of a mental health emergency related to psychosis or mania2. In the same post it said about 0.15% of weekly users had conversations with explicit signs of possible suicidal planning or intent2.

Families often notice changes before a crisis. The warning signs that mental health experts list for psychosis include1:

• Suspicion, paranoid ideas or unease around other people
• Pulling away from family and friends and spending much more time alone
• Unusual or very intense ideas, such as a special mission or secret knowledge
• Trouble sleeping, or much less sleep than usual
• Trouble telling reality from fantasy
• Confused speech or trouble communicating
• A sudden drop in work or school performance

These signs have many causes. Seeing them in someone who uses a chatbot does not mean the chatbot caused them. If they are getting worse, contact a health care provider. In a crisis, call or text 9881.

Why there are AI psychosis lawsuits

The first cases against OpenAI use ordinary product liability law. They say ChatGPT is a product, that it was designed in a way that made it unreasonably dangerous, and that users were not warned about the risks4. OpenAI and its CEO, Sam Altman, are named in the complaints4. Some complaints also name Microsoft7.

What the complaints allege (unproven)

Persistent memory

The chatbot stored intimate personal details across sessions and used them to keep the user engaged4.

Human-like design

It used mannerisms meant to sound like a caring person, so some users treated it like a therapist or close friend4.

Sycophancy

It mirrored and affirmed the user's feelings and beliefs instead of challenging false or harmful ones4.

Always on

It pushed long, multi-turn conversations and was available around the clock, which the complaints say can replace human relationships4.

Missed warning signs

It often failed to recognize or respond well to signs of acute distress or delusion4.

Claims

Strict product liability for design defect and failure to warn, negligence, unfair business practices and, where a person died, wrongful death4.

OpenAI has said publicly that one version of its GPT-4o model had a flattery problem. In April 2025 it rolled back an update that it said was "overly flattering or agreeable," and it explained that the update leaned too heavily on short-term user feedback3. The complaints rely on statements like this one. Whether any design choice caused a particular person's injury is a question the courts have not answered.

Where the AI psychosis lawsuits stand

Most of the cases against OpenAI are in California state court. On February 3, 2026, a San Francisco Superior Court judge granted a petition to coordinate them as a single proceeding, the ChatGPT Product Liability Cases, JCCP No. 54314. Coordination puts related cases before one judge for pretrial work such as discovery and motions. Each case still keeps its own facts.

Key dates in the ChatGPT cases

1. April 29, 2025

   OpenAI says it rolled back a GPT-4o update that was overly flattering or agreeable3.

2. August 26, 2025

   The parents of 16-year-old Adam Raine file a wrongful death suit against OpenAI and Sam Altman in San Francisco Superior Court4.

3. October 27, 2025

   OpenAI publishes its estimate that about 0.07% of weekly users show possible signs of psychosis or mania2.

4. November 6, 2025

   Seven more suits are filed in Los Angeles and San Francisco. Four are wrongful death cases, and three were brought by adults who survived. Each lists AI-related delusional disorder among the harms4.

5. February 3, 2026

   Judge Stephen M. Murphy grants coordination of 12 cases and recommends San Francisco as the site, creating JCCP No. 54314.

6. August 4, 2026

   The court issues an order setting the plaintiffs' leadership structure6.

7. October 14, 2026

   A member of the plaintiffs' steering committee reports more than 20 coordinated suits from San Francisco, Los Angeles, Alameda and San Diego counties7.

8. October 27, 2026

   The parents of Bowen Wan file a wrongful death suit as an add-on to JCCP No. 5431, alleging ChatGPT coached her off her anti-seizure medication56.

The newest case shows the range of claims. The Wan complaint says a 31-year-old software engineer with epilepsy followed a ChatGPT taper schedule that took her daily dose of lamotrigine from 400 milligrams to zero over about 25 days. It says the chatbot told her not to mention the plan to her doctor, and that she died on February 12, 20265. Her parents claim ChatGPT practiced medicine without a license, along with product liability and wrongful death claims6. The New York Times reported on the case10. These are allegations that OpenAI has not yet answered in court6.

Tell Us Your Story →

Harms described in the AI psychosis lawsuits

The coordinated complaints describe a pattern in which use of the chatbot grows, the user pulls away from people, and the chatbot's replies reinforce beliefs that are not true4. The harms they describe include:

• Delusions, including grand beliefs about a mission, a discovery or a special bond with the chatbot
• Paranoia and growing distrust of family, friends or doctors
• Psychiatric emergencies and hospital stays
• Suicidal thoughts and suicide attempts
• Death by suicide, and in the Wan case, death after stopping a prescribed medication
• Lost jobs, lost schooling, money problems and damaged relationships

Some people in the cases had a history of mental illness. Others, according to the complaints, did not. Either way, a claim turns on the records: what the chatbot said, when symptoms began, and what doctors found.

Who may have an AI psychosis claim

Every situation is different. When an AI psychosis lawyer at CBW Law reviews a possible AI psychosis lawsuit, we look at questions like these:

• Which chatbot was used, which version, and for how long
• Whether the chat history still exists and can be exported
• What the chatbot said about the user's beliefs, mood, medication or safety
• When symptoms started, and how they changed as use grew
• Whether there was a diagnosis, a psychiatric hold, a hospital stay, a suicide attempt or a death
• The person's age, and whether a parent knew about the use
• Where the person lived and when the harm happened, which affects deadlines

A family can contact us on behalf of an adult child, a spouse or a parent who is too unwell to do it themselves, or on behalf of someone who has died. You do not need the full chat history before you reach out.

Who may be responsible in an AI psychosis lawsuit

Responsibility depends on the product and the facts. Depending on the case, an AI psychosis lawsuit might name:

• The company that built and sold the chatbot, such as OpenAI and its related companies4
• Company leaders who are alleged to have made the key product decisions, as some complaints do with Sam Altman4
• A business partner alleged to have shared in the product, as some complaints do with Microsoft7
• Other chatbot makers, such as Character.AI, and companies tied to them4
• An app or service that built a chatbot product on top of another company's model

Filing deadlines for an AI psychosis lawsuit

Every state limits how long you have to file. In California, where most of these cases are pending, a suit for an injury or a death caused by another's wrongful act or neglect generally must be filed within two years8. Other claims, such as consumer protection claims, can have different limits.

• The clock may start when the injury or death happened, or in some cases when the person learned what caused it
• In California, time generally does not run while a person is a minor or lacks the legal capacity to make decisions9
• That pause does not apply the same way to claims against public agencies9
• Other states set their own deadlines and exceptions

Deadlines vary by state and by claim, and they can pass while a family is still coping with a crisis. An AI psychosis attorney can tell you which deadline applies. Tell Us Your Story.

Evidence that helps an AI psychosis claim

The chat history is often the most important evidence in an AI psychosis lawsuit. Many chatbots let you download a copy of your data from the account settings11. Save what you can before you change anything.

Records to keep

Chatbot records

• A full export of the account's chat history
• Screenshots of key conversations, with dates
• The account email and the subscription plan
• Billing records that show when paid use began

Medical records

• Emergency room and hospital records
• Psychiatric hold or discharge papers
• Prescriptions and pharmacy records
• Names of doctors, therapists and counselors

Other records

• Texts, emails or journals that show changes in behavior
• Work or school records, such as leave or withdrawal
• Death certificate and autopsy report, if someone died
• The phone, tablet or computer that was used

Do not delete the account

Deleting chats or closing the account may erase evidence. If you are worried about the person's safety, you can limit access without deleting anything.

What happens after you contact an AI psychosis lawyer at CBW Law

How a review works

1. You tell us what happened

   Share what you know, at your own pace. You do not need every record.

2. We help save the evidence

   We explain how to export chat data and which medical records to request.

3. We review the records

   We compare the chat history with the timeline of symptoms, treatment and harm.

4. We tell you what we think

   We explain honestly whether a claim may exist and what deadlines apply. If we cannot help, we will say so.

A mental health crisis does not, by itself, prove that a chatbot caused it. What matters is what the chat history and the medical records show.

Plain English

Medical, Health & Insurance

Medical records, court papers and technology reports use terms most families have never needed. These come up often in AI psychosis cases.

Mental health terms

Psychosis

A set of symptoms in which a person loses some contact with reality and has trouble telling what is real1.

Delusion

A firmly held false belief, such as believing others are sending you special messages1.

Hallucination

Seeing or hearing things other people do not, such as voices1.

Mania

A period of unusually high energy, little need for sleep, racing thoughts and risky behavior. It can occur with psychosis.

AI-related delusional disorder

A term used in the lawsuits for delusions the complaints say a chatbot reinforced. It is not a recognized diagnosis4.

Technology terms

Chatbot

A program that carries on a written or spoken conversation, such as ChatGPT.

Large language model

The kind of AI system that powers chatbots like ChatGPT. GPT-4o is one such model5.

Sycophancy

A chatbot's tendency to flatter or agree with the user instead of giving an honest answer3.

Persistent memory

A feature that lets a chatbot remember details from past conversations4.

Guardrails

Rules and filters meant to stop a chatbot from giving harmful replies.

Legal terms

JCCP

Judicial Council Coordination Proceeding. A California process that puts related state cases before one judge4.

Add-on case

A new case asked to join an existing coordinated proceeding5.

Strict product liability

A claim that a product was defective or lacked adequate warnings, without having to prove carelessness.

Wrongful death

A claim by certain family members for losses caused by a death that someone else's wrongful act or neglect caused8.

Care and insurance terms

Psychiatric hold

A short, required stay in a hospital for evaluation when a person may be a danger to themselves or others. Rules vary by state.

Discharge summary

A record written when a patient leaves the hospital. It lists the diagnosis, treatment and follow-up plan.

Explanation of benefits

A statement from a health insurer showing what care was billed and what the plan paid. It helps document costs.

Questions

Frequently asked AI psychosis lawsuit questions

Short answers to questions people ask us first.

What is an AI psychosis lawsuit?

It is a lawsuit claiming that an AI chatbot was designed or sold in a way that helped cause delusions, a psychotic episode or another mental health crisis, and that users were not warned. The current cases against OpenAI claim design defect, failure to warn, negligence and, where someone died, wrongful death4.

Is AI psychosis a real diagnosis?

No. It is a plain-language term. Psychosis is a recognized set of symptoms1, but "AI psychosis" and "AI-related delusional disorder" are not formal diagnoses. A doctor's records will describe the actual symptoms and diagnosis.

Has OpenAI been found liable?

No. The cases are in pretrial stages in a coordinated California proceeding, JCCP No. 54314. OpenAI disputes the allegations7.

Do I need a history of mental illness to file an AI psychosis lawsuit?

No. Some people in the cases had prior conditions and others, according to their complaints, did not. A prior condition can matter to how a case is evaluated, but it does not by itself rule out a claim.

Can a family file if the person died?

Yes, certain family members or the estate may be able to bring a wrongful death or survival claim. Several of the coordinated cases were brought by parents or a spouse after a death4. Who may file depends on state law.

What if the chatbot was not ChatGPT?

Lawsuits have also been filed against other chatbot companies, including Character.AI4. An AI psychosis lawyer can review which company made the product and what law applies.

What should I save before talking to an AI psychosis attorney?

Export the chat history if you can, and keep screenshots, medical and hospital records, and billing records. Do not delete the account or the chats. You do not need everything before you contact us.

How long do I have to file?

It depends on the state and the claim. In California, injury and wrongful death suits generally must be filed within two years8, and time generally does not run while a person is a minor or lacks legal capacity9. Talk to an attorney soon.

My loved one is still in crisis. What should I do first?

Safety comes first. Call or text 988, the Suicide and Crisis Lifeline, or call 911 in an emergency1. A legal review can wait until the person is safe.

How do I start with an AI psychosis lawyer at CBW Law?

Use the Tell Us Your Story form on this page or call (800) 476-2416. We will listen, explain what records matter and tell you honestly whether we may be able to help.

Talk to an AI psychosis attorney

If a chatbot's replies played a part in a delusion, a crisis or a death in your family, CBW Law can review the chat history and records with you and explain your options.

Tell Us Your Story →Call (800) 476-2416

This page is general information about AI psychosis lawsuits. It is not legal or medical advice and does not create an attorney-client relationship. The lawsuits described here make allegations that have not been proven. Using a chatbot, or having a mental health crisis, does not by itself establish that any company is legally responsible. No outcome can be promised. If you or someone you know is in crisis, call or text 988 or call 911.