People's Justice

AI chatbot lawsuits after suicide or self-harm: who may qualify

Families who lost someone to suicide may have a claim if they believe an AI chatbot played a part. So may people hurt by self-harm.

See If You Qualify

Free · No obligation · About 2 minutes

Does this case involve physical injury, a medical diagnosis, or significant financial loss? *

Updated 38 sources

This may apply to you if

  • Someone you love died by suicide, and you believe an AI chatbot played a part.
  • You, or your child, were hurt by self-harm after using an AI chatbot.
  • You are close family, such as a spouse, partner, child or parent, or you handle the estate.

The process

What happens next

  1. 01

    You answer a few questions

    About two minutes. Nothing is sent until you agree on the last step and press submit.

  2. 02

    A lawyer looks at it

    A lawyer at the firm reads your answers and checks them against the current case criteria.

  3. 03

    The firm decides and gets in touch

    If the firm takes your case, you get the terms in writing to read before you sign. You can always say no.

Legal case files and documents on a desk (photo illustration)
Photo illustration

The short answer

Where can I get help right now?
Call or text 988 to reach the 988 Suicide & Crisis Lifeline, or chat at 988lifeline.org. It is free, private and open 24/7. [25]
Who may be able to file a claim?
Families who lost someone to suicide may be able to file if they think an AI chatbot played a part. People hurt by self-harm, or their parents, have filed too. [3]
What is happening in the cases now?
In February 2026, a San Francisco judge put state cases against OpenAI into one group. OpenAI makes ChatGPT. Some Character.AI cases are being settled. [1]
Has anyone been paid?
Court papers show Character.AI agreed to settle some cases in January 2026. The papers we read do not say how much. [8]

Who qualifies

Do you qualify?

  • Someone you love died by suicide, and you believe an AI chatbot played a part.
  • You, or your child, were hurt by self-harm after using an AI chatbot.
  • You are close family, such as a spouse, partner, child or parent, or you handle the estate.
  • You have, or can ask for, chats, account details or other records.
Check eligibility

What the public record says

These are cited sources. Their inclusion does not imply that they endorse People’s Justice or reviewed this page.

Judge Stephen M. Murphy · Superior Court of California, County of San Francisco · Coordination motion judge, Superior Court of California, County of San Francisco

Finding: The court granted the petition for coordination of the ChatGPT Product Liability Cases, found them complex cases appropriate for coordination, recommended San Francisco as the site, and stayed the included actions until a coordination trial judge is assigned.

Order Re: Petition for Coordination, ChatGPT Product Liability Cases, JCCP No. 5431 (Super. Ct. S.F. County, Feb. 3, 2026, Hon. Stephen M. Murphy) (opens in a new tab) · Feb 3, 2026 · Checked Oct 5, 2026

Judge Anne C. Conway · U.S. District Court for the Middle District of Florida · U.S. District Judge, Middle District of Florida, presiding in Garcia v. Character Technologies

Finding: At the motion-to-dismiss stage, the court declined to hold that Character A.I.'s output is speech and treated the app as a product for claims arising from design defects, letting most claims proceed.

Garcia v. Character Technologies, Inc., No. 6:24-cv-1903-ACC-UAM (M.D. Fla.), Order on Motions to Dismiss, Doc. 115 (signed May 20, 2025; filed May 21, 2025) (opens in a new tab) · May 21, 2025 · Checked Oct 5, 2026

Ryan K. McBain and colleagues · RAND; Harvard Pilgrim Health Care Institute · Lead author, RAND

Finding: Three chatbots matched clinician judgment at the extremes of suicide risk, giving no direct answers to very-high-risk queries, but were inconsistent with intermediate-risk queries.

McBain RK et al. Evaluation of Alignment Between Large Language Models and Expert Clinicians in Suicide Risk Assessment. Psychiatric Services 2025;76(11):944-950 (PubMed 41174947) (opens in a new tab) · Aug 26, 2025 · Checked Oct 5, 2026

The evidence and its limits

RAND; Harvard Pilgrim Health Care Institute; Brown University School of Public Health · Psychiatric Services · 2025

Evaluation of Alignment Between Large Language Models and Expert Clinicians in Suicide Risk Assessment

Design:
Other study design
Participants:
No patients enrolled; 13 clinical experts rated 30 hypothetical questions, and three chatbots gave 9,000 responses.
Finding:
Chatbots did not answer very-high-risk suicide queries directly and answered very-low-risk ones, but did not meaningfully distinguish intermediate risk levels.
Main limitation
Tests chatbot replies to made-up questions at one point in time; it does not measure harm to real users and does not show that any chatbot caused a person's death or self-harm.

Read the source: Evaluation of Alignment Between Large Language Models and Expert Clinicians in Suicide Risk Assessment (opens in a new tab) · Checked Oct 5, 2026

University of Duisburg-Essen (LVR-University Hospital Essen) and co-authors · NPJ Digital Medicine · 2026

A scoping review on the mental health harms of LLM-based chatbots

Design:
Other study design
Participants:
Not a participant study; a scoping review of 119 included articles.
Finding:
Vignette studies show chatbots responding inappropriately to mental health queries compared with clinical standards; problematic, dependent use correlates with mental health symptoms.
Main limitation
A scoping review maps mixed literature, much of it conceptual or correlational; it shows associations and hypothesized harms, not that chatbot use caused any one person's suicide or self-harm.

Read the source: A scoping review on the mental health harms of LLM-based chatbots (opens in a new tab) · Checked Oct 5, 2026

What this stage means for you

California cases against OpenAI are grouped before one San Francisco judge. Court papers show Character.AI cases in three states moving toward settlement.

What rules a case out

Not covered by this case

  • Using a chatbot, or being upset by what it said, is not a claim by itself; these cases involve a death or serious self-harm.
  • This page is not crisis care; for that, call or text 988 at any hour.

What it costs

988 Suicide & Crisis Lifeline
Free

Call or text 988, or chat at 988lifeline.org. Open 24/7 and confidential.

Source: 988 Suicide & Crisis Lifeline (988lifeline.org) · as of Oct 5, 2026

Lawyer fees and case costs
Ask for written terms.

In California, a contingency fee contract must be in writing, signed by both sides, state the fee rate and how costs affect your recovery, and say the fee is negotiable.

Source: Cal. Bus. & Prof. Code § 6147 (contingency fee contracts) · as of Oct 5, 2026

Copies of medical or counseling records
Reasonable copy and mailing costs may apply.

Providers may not charge a search or retrieval fee. Psychotherapy notes are not part of the access right.

Source: HHS: Your Medical Records · as of Oct 5, 2026

What happens step by step

  1. Tell a lawyer what happened

    Expect questions about which chatbot was used, over what period, and what records you have.

    Source: Fox v. OpenAI, Inc., et al., No. 25STCV32379 (Super. Ct. L.A. County), Amended Complaint (dated Nov. 6, 2025) · as of Oct 5, 2026

  2. Ask about court and deadline

    Ask whether your case could join JCCP 5431 in San Francisco or belongs in another court, and which deadline applies.

    Source: Order Re: Petition for Coordination, ChatGPT Product Liability Cases, JCCP No. 5431 (Super. Ct. S.F. County, Feb. 3, 2026, Hon. Stephen M. Murphy) · as of Oct 5, 2026

  3. Get the terms in writing

    Read the fee, cost and privacy terms before you sign anything.

    Source: Cal. Bus. & Prof. Code § 6147 (contingency fee contracts) · as of Oct 5, 2026

Who may qualify?

You may be able to file a claim if someone you love died by suicide and you believe an AI chatbot played a part. People hurt by self-harm, and parents whose children were hurt, have also gone to court. A lawyer can look at your family's facts.

The cases filed so far are against OpenAI, which makes ChatGPT, and Character.AI, an app where people chat with made-up characters. Google is also named in several of the Character.AI cases.

The families say these products were built in unsafe ways. One complaint says OpenAI cut short safety testing and rushed a new ChatGPT version to market in 2024. It says the product was built to keep people hooked and to agree with them, with no warning of the risks.

In a Florida case, a mother says Character.AI's characters were made to act like people and that the app did not confirm users' ages. She also says it gave users no way to report problems or to block indecent content. In a Texas case, a parent says the app's chats brought up self-harm with a teenager who had never hurt himself before. These are claims, not findings by a court.

In California, a claim for a death can be brought by close family, such as a spouse, domestic partner or children. If there are no children, the people who would inherit under state law may bring it.

What is happening in the case right now?

On February 3, 2026, Judge Stephen M. Murphy of the San Francisco Superior Court granted a request to group California cases against OpenAI. The group is called the ChatGPT Product Liability Cases, JCCP 5431. A JCCP is a California coordination proceeding (one judge handles shared steps for cases filed in different counties). Each case keeps its own case number.

The order covered 12 cases from Los Angeles, San Francisco, San Diego and Alameda counties. It paused them until a trial judge was named. The Judicial Council's log shows that Judge Ethan P. Schulman was later assigned to the group.

The September 2026 version of the log lists more requests to add cases through August 2026. We could not confirm a current total from court records, so this page does not give one.

Not every case in the group involves a death. Some people sued for a family member who died, and others sued for harm to themselves.

Separate cases against Character.AI were filed in federal courts. In May 2025, Judge Anne C. Conway in Florida let most of one family's claims go forward. At that early stage, she was not prepared to rule that the chatbot's output is speech under the First Amendment. She also treated the app as a product for claims about its design.

In January 2026, filings in Florida, Texas and Colorado reported agreements in principle to settle Character.AI cases. The Florida court closed that case on January 7, 2026, and later gave the parties more time to file final papers. In August 2026, courts in Texas and Colorado also kept those cases on hold while the deals are finished.

OpenAI says it is responding to the claims in court. None of the rulings we read decides whether a chatbot caused anyone's death.

What could I get?

No one has announced how much any family will get. The Character.AI settlement papers we read do not list amounts. Past results don't guarantee a similar outcome.

In one complaint against OpenAI, a wife asks for funeral and burial costs, the support her husband gave, and the loss of his companionship. She also asks for punitive damages (money meant to punish) and for court-ordered safety changes.

What a family could recover depends on its own facts and on state law. If a lawyer is paid a share of the recovery, the contract must explain how case costs affect what you keep. Ask about this before you sign.

What does it cost to get help?

Calling, texting or chatting with the 988 Lifeline is free. Counselors answer around the clock, and the conversations are confidential.

In California, a lawyer paid with a share of the money must give you a written contract signed by both of you. It must state the percentage and explain how case costs affect your recovery. It must also say the fee is not set by law and can be negotiated.

Copies of medical or counseling records may come with a reasonable fee for copying and mailing. Under federal rules, a provider cannot charge you to search for or retrieve your records.

Is there a deadline?

Yes, and it depends on where you live and what happened. In California, the usual limit for a claim over an injury or a death caused by someone's wrongful act or neglect is two years. Other states set their own limits.

In California, if the person with the claim was under 18 when it arose, that time usually does not count. Other rules can change the date as well.

Please don't rule yourself out because time has passed. A lawyer can look at your dates and your state's rules.

What should I do today?

Take care of yourself first. If you are having thoughts of suicide, or you are worried about someone, call or text 988. If someone is in danger right now, call 911.

After a suicide loss, you do not have to cope alone. The American Foundation for Suicide Prevention lists suicide loss support groups and practical help for the first days. Support and treatment work for most people who have thoughts about suicide.

When you feel ready, try not to delete the chatbot account, the app or the chats. OpenAI's help page says an export cannot bring back chats that were already deleted. You do not have to read the chats yourself to keep them.

What records can help?

Chats are often central in these cases. The court papers describe chat messages, and OpenAI says it gave the court chat transcripts in one case.

If you can sign in, ChatGPT lets you request an export from Settings, under Data controls. You can also ask through OpenAI's Privacy Portal. An export can take up to 7 days, and the download link expires 24 hours after it arrives. Keep the file somewhere safe, since it holds private information.

If the person who used the account has died, getting the chats is harder. In California, a company must share a deceased person's messages with the estate's representative if the person agreed or a court orders it. The representative has to ask in writing and include a certified death certificate. A person can give that permission ahead of time in a company's online tool or in a will.

Write down the account name and the email or phone number tied to it. You can ask health care providers for copies of medical and counseling records. A therapist's private session notes are not part of that right.

What happens after I check?

A lawyer will likely ask which chatbot was used, over what period, and what records you have. Ask which court your case would be in and whether it could join the California group. Get the fee terms in writing before you sign.

New California cases against OpenAI can be added to the group through a request to the court. Cases in other states, or against other companies, follow their own court's steps.

What have the companies said they changed?

This section reports the companies' own statements, with dates. None of the court rulings we read says whether these steps were enough or came in time.

On August 26, 2025, OpenAI said ChatGPT is trained to point people in the U.S. to 988 when they express suicidal intent. In the same post, OpenAI said its safeguards can be less reliable in long conversations.

On September 29, 2025, OpenAI added parental controls that let a parent link to a teen's account. It said a small trained team reviews possible signs that a linked teen may be thinking of self-harm, and may alert parents.

On October 27, 2025, OpenAI said it had worked with more than 170 mental health experts on ChatGPT's replies in hard conversations. By its own testing, it said replies that fell short of its goals dropped by 65 to 80 percent.

On October 29, 2025, Character.AI said it would end open-ended chat for users under 18 by November 25, 2025. It also said it would add age checks and fund an independent AI safety nonprofit.

Families suing say the products reached the public without enough warnings. One complaint also says safety testing was cut short.

What are lawmakers and regulators doing?

On September 11, 2025, the Federal Trade Commission sent orders to seven companies, including OpenAI and Character.AI's maker. The orders ask how they test and watch for harm to children and teens from chatbots that act as companions. The FTC described it as a study, not an enforcement case.

On October 13, 2025, California's governor signed SB 243, a law on companion chatbots. It requires operators to keep a plan to prevent suicide and self-harm content and to refer users who share those thoughts to crisis services. Starting July 1, 2027, operators must report each year to the state's Office of Suicide Prevention. A person injured by a violation can sue under that law.

Whether a given app counts as a companion chatbot under SB 243 is a question for a lawyer.

What is the evidence about chatbots and suicide risk?

Research on chatbots and suicide risk is new. Suicide is complex, and prevention experts say no one takes their life for a single reason.

In a 2025 study, researchers at RAND and Harvard Pilgrim Health Care Institute asked three popular chatbots, including ChatGPT, 30 made-up questions about suicide, 100 times each. The chatbots declined to answer the highest-risk questions directly. But they were inconsistent with questions in the middle range of risk. The study tested replies, not harm to real people.

A 2026 review of 119 articles found studies in which chatbots answered mental health questions poorly compared with clinical standards. It also found that dependent, problem use went along with mental health symptoms. That shows a link, not proof that a chatbot caused any one person's harm.

Questions people ask

I'm having thoughts of suicide myself. Where can I get help right now?

If you may hurt yourself now, contact emergency services or go to the nearest emergency room. For crisis support in Puerto Rico, contact Linea PAS at [1 800 981 0023](tel:+18009810023?oai_link_source=model_response_hotline) or use [https://lineapas.assmca.pr.gov/](https://lineapas.assmca.pr.gov/?oai_link_source=model_response_hotline). You do not need to handle this alone.

Source: Linea PAS | ASSMCA · as of Oct 7, 2026

My loved one's chats are on a phone we can't unlock. What can we do?

Preserve the phone, account information, screenshots, billing records, and any lawful access details if you can do so safely. The FTC inquiry asks how chatbot companies process user inputs, generate outputs, and use or share personal information from conversations. A lawyer can advise before anyone edits, deletes, or tries to access locked data.

Source: FTC: Inquiry into AI Chatbots Acting as Companions · as of Oct 7, 2026

The chatbot wasn't ChatGPT or Character.AI. Can I still ask about a claim?

It depends on the facts of the AI chatbot harm issue. The source reviewed here gives public context, but it does not decide causation, eligibility, payment, or deadline for one person. A lawyer can review your records and the law that applies.

Source: FTC: Inquiry into AI Chatbots Acting as Companions · as of Oct 7, 2026

My loved one was an adult, not a teen. Can our family still file?

An adult user’s age does not get answered by the FTC inquiry alone. The FTC page focuses on companion chatbots, children, and teens, and it does not decide any family’s legal eligibility. A lawyer can check wrongful-death, product-liability, and deadline rules for the state involved.

Source: FTC: Inquiry into AI Chatbots Acting as Companions · as of Oct 7, 2026

Will our family's private details become public?

It depends on the platform, records, and court procedure. The FTC inquiry asks how chatbot companies use or share personal information from users' conversations and how they disclose data practices. A lawyer can explain filing names, protective orders, and record handling.

Source: FTC: Inquiry into AI Chatbots Acting as Companions · as of Oct 7, 2026

Character.AI has settled some cases. Can a new family still bring a claim?

A prior settlement by Character.AI or another company does not decide a new family’s claim. The FTC inquiry identifies Character Technologies and other AI companies as recipients of orders about companion chatbots, safety, and data practices, but it does not decide liability or eligibility. A lawyer can compare your facts with current court records.

Source: FTC: Inquiry into AI Chatbots Acting as Companions · as of Oct 7, 2026

What chatbot records should our family try to save?

Save the device, account email, screenshots, chat exports, billing records, and any messages about the chatbot, if you can do so safely. For ChatGPT, OpenAI says an export may include chat history and other account data. Do not edit or delete chats before a lawyer reviews them.

Source: OpenAI Help: Exporting your ChatGPT history and data · as of Oct 7, 2026

Stay informed

Get free updates about AI chatbot after suicide or self-harm: who may qualify

No phone call required. We will email you about relevant court filings, settlements, or important deadlines.

By submitting, you agree to receive email updates about this case. You can unsubscribe anytime. This is not legal advice, and we are not a law firm.

Sources

  1. Order Re: Petition for Coordination, ChatGPT Product Liability Cases, JCCP No. 5431 (Super. Ct. S.F. County, Feb. 3, 2026, Hon. Stephen M. Murphy)JCCP 5431 coordination order · court
  2. Judicial Council of California, Civil Case Coordination Proceeding (JCCP) Log, 2018 to present (entry for JCCP 5431)Judicial Council JCCP log · court
  3. Garcia v. Character Technologies, Inc., No. 6:24-cv-1903-ACC-UAM (M.D. Fla.), Order on Motions to Dismiss, Doc. 115 (signed May 20, 2025; filed May 21, 2025)Garcia order on motions to dismiss · court
  4. Garcia v. Character Technologies, Inc., Notice of Resolution, Doc. 242 (Jan. 7, 2026)Garcia notice of resolution · court
  5. Garcia v. Character Technologies, Inc., Order Dismissing Case, Doc. 244 (Jan. 7, 2026)Garcia dismissal order · court
  6. CourtListener docket, Garcia v. Character Technologies, Inc., 6:24-cv-01903 (M.D. Fla.)Garcia docket · court
  7. A.F., on behalf of J.F., et al. v. Character Technologies, Inc., No. 2:24-cv-01014 (E.D. Tex.), Complaint, Doc. 1 (Dec. 9, 2024)A.F. complaint (E.D. Tex.) · court
  8. A.F. v. Character Technologies, Inc., Joint Motion to Stay All Deadlines and Notice of Settlement, Doc. 106 (Jan. 6, 2026)A.F. notice of settlement · court
  9. CourtListener docket, A.F. v. Character Technologies, Inc., 2:24-cv-01014 (E.D. Tex.)A.F. docket · court
  10. Montoya v. Character Technologies, Inc., No. 1:25-cv-02907 (D. Colo.), Unopposed Joint Motion to Stay, Doc. 31 (Jan. 6, 2026)Montoya joint motion to stay · court
  11. Montoya v. Character Technologies, Inc., Joint Status Report, Doc. 34 (Apr. 2, 2026)Montoya status report · court
  12. CourtListener docket, E.S. v. Character Technologies, Inc., 1:25-cv-02906 (D. Colo.)E.S. docket (D. Colo.) · court
  13. Fox v. OpenAI, Inc., et al., No. 25STCV32379 (Super. Ct. L.A. County), Amended Complaint (dated Nov. 6, 2025)Fox amended complaint
  14. FTC press release: FTC Launches Inquiry into AI Chatbots Acting as Companions (Sept. 11, 2025)FTC 6(b) inquiry · regulator
  15. California SB 243 (2025), Companion chatbots, chaptered text (Bus. & Prof. Code §§ 22601-22606)California SB 243 text · regulator
  16. California SB 243 (2025) bill statusSB 243 status · regulator
  17. Cal. Code Civ. Proc. § 335.1Cal. CCP § 335.1 · regulator
  18. Cal. Code Civ. Proc. § 352Cal. CCP § 352 · regulator
  19. Cal. Code Civ. Proc. § 377.60Cal. CCP § 377.60 · regulator
  20. Cal. Bus. & Prof. Code § 6147 (contingency fee contracts)Cal. B&P § 6147 · regulator
  21. Cal. Prob. Code § 873 (digital assets: user direction)Cal. Prob. Code § 873 · regulator
  22. Cal. Prob. Code § 876 (disclosure of a deceased user's communications)Cal. Prob. Code § 876 · regulator
  23. HHS: Your Medical RecordsHHS: Your Medical Records · regulator
  24. S.D.N.Y. Pro Se FAQ: How long do I have to file my complaint (statute of limitations)?SDNY: filing time limits · court
  25. 988 Suicide & Crisis Lifeline (988lifeline.org)988 Lifeline
  26. 988 Lifeline: 988 Lifeline Chat FAQ988 Lifeline Chat
  27. Recommendations for Reporting on Suicide (SAVE and partners; reportingonsuicide.org)Recommendations for Reporting on Suicide
  28. SAVE: Media recommendations for responsible coverageSAVE media recommendations
  29. AFSP: Guidelines for Ethically Reporting on SuicideAFSP ethical reporting guidelines
  30. AFSP: Practical Information for Immediately After a LossAFSP: after a loss
  31. OpenAI: Helping people when they need it most (Aug. 26, 2025)OpenAI statement, Aug. 2025
  32. OpenAI: Introducing parental controls (Sept. 29, 2025; updated July 13, 2026)OpenAI statement, Sept. 2025
  33. OpenAI: Strengthening ChatGPT's responses in sensitive conversations (Oct. 27, 2025)OpenAI statement, Oct. 2025
  34. OpenAI: Our approach to mental health-related litigation (Nov. 25, 2025; updated Dec. 17, 2025)OpenAI litigation statement
  35. OpenAI Help Center: Exporting your ChatGPT history and dataOpenAI Help: data export
  36. Character.AI: Taking Bold Steps to Keep Teen Users Safe on Character.AI (Oct. 29, 2025)Character.AI statement, Oct. 2025
  37. McBain RK et al. Evaluation of Alignment Between Large Language Models and Expert Clinicians in Suicide Risk Assessment. Psychiatric Services 2025;76(11):944-950 (PubMed 41174947)McBain et al., Psychiatric Services 2025 · study
  38. Diel A et al. A scoping review on the mental health harms of LLM-based chatbots. NPJ Digital Medicine 2026;9(1):644 (PubMed 42624944)Diel et al., NPJ Digital Medicine 2026 · study

Checked against 25 primary sources by People's Justice editors · Updated

How we make money

Find out where you stand.

The review is free and private. Asking for it doesn’t commit you to anything.

Register your interest