Published
September 17, 2026
OpenAI faces roughly two dozen coordinated product-liability cases in California, a first-in-the-nation state lawsuit from Florida, and a subpoena from a coalition of state attorneys general — but it has never paid a tort claim to a consumer or a state. Whether that changes before 2028 depends on three separate legal tracks, and they are not moving at the same speed.
The Question, Defined Narrowly
"Will OpenAI pay?" is too loose to answer. OpenAI is a defendant in copyright suits, contract disputes, trademark fights, privacy cases and a fast-growing set of personal-injury actions, and those categories pay out on completely different schedules and for completely different reasons. So this analysis asks a narrower question with a hard edge on it:Will money change hands, on or before December 31, 2027, on a claim sounding in tort — personal injury, product liability, negligence, or consumer protection — paid either to injured individuals, to a certified consumer class, or to a state?
That definition deliberately excludes three things people often count as "OpenAI paying." It excludes copyright and licensing money, which is contract and intellectual-property exposure rather than injury compensation. It excludes defense costs and regulatory compliance spending, which are large but are not damages. And it excludes any payment whose existence is only rumored — the question is about money that can be documented.
What OpenAI Has and Has Not Paid So Far
As of September 17, 2026, there is no public record of OpenAI paying damages on a personal-injury, product-liability or consumer-protection claim, and no court has entered a damages judgment against it on one. No consumer class has been certified against the company in a tort case. There is no OpenAI settlement fund, no administrator, no claim form and no deadline.That matters because the AI sector has already produced a very large payout — just not from OpenAI, and not in tort. Anthropic agreed to a $1.5 billion settlement of copyright claims brought by authors over books used to train Claude, and a federal court granted it final approval on July 20, 2026. That number gets quoted as though it sets a price for AI harm generally. It does not. Copyright carries statutory damages that a court can compute per infringed work without anyone proving they were hurt, which is why that case could be valued and resolved in roughly two years. A tort claim requires a plaintiff to prove that a particular person suffered a particular injury and that the product caused it. Those cases take longer, settle later, and are valued one at a time.
See OCA's coverage of the Anthropic books copyright settlement and its final approval for how that one actually worked.
Track One — The Personal-Injury Docket
The most likely source of an OpenAI tort payment before 2028 is not a class action. It is the set of individual suits now coordinated in California.In August 2025, the parents of a 16-year-old who died by suicide sued OpenAI and Sam Altman in San Francisco Superior Court in Raine v. OpenAI, alleging that ChatGPT deepened their son's suicidal ideation rather than interrupting it. The complaint was amended that October to add allegations of intentional misconduct, and OpenAI responded in November 2025 disputing causation. Similar suits followed from other families and from adults alleging psychological harm.
On February 3, 2026, the San Francisco branch of the California Superior Court entered a coordination order creating Judicial Council Coordination Proceeding No. 5431, captioned In re ChatGPT Product Liability Cases. It started with roughly a dozen cases and has since grown to about two dozen, drawn from San Francisco, Los Angeles, Alameda and San Diego counties with plaintiffs from around the country. The proceeding is assigned to Judge Ethan P. Schulman. On August 4, 2026, he entered Case Management Order No. 1, appointing plaintiffs' co-lead counsel, a liaison counsel and a five-member steering committee. The next case management conference is set for September 23, 2026.
Plaintiffs in these cases allege, among other things, that ChatGPT is unreasonably dangerous as designed, that it reinforced delusional thinking, that it failed to interrupt conversations about self-harm, and that OpenAI weakened safety guardrails in pursuit of engagement. OpenAI disputes the allegations and denies that its product caused these injuries. None of the allegations has been proven, and no case in JCCP 5431 has been tried.
A JCCP is a pretrial coordination device, not a class action. There is no class, no class members, nobody to opt in, and nothing to claim. Each plaintiff keeps their own case; what gets shared is discovery, motions and a judge. That structure is what makes an early payment plausible: a defendant can resolve a handful of individual cases quietly, without a certified class, without a public fund, and without any finding of liability.
Track Two — The States
On June 1, 2026, Florida became the first state to sue an AI company. Attorney General James Uthmeier filed a ten-count civil complaint in the Circuit Court of the Tenth Judicial Circuit in Highlands County against five OpenAI entities and against Altman personally.The counts run through the Florida Deceptive and Unfair Trade Practices Act — unfair, unconscionable and deceptive practices, plus a FDUTPA count built on alleged non-compliance with the federal Children's Online Privacy Protection Act — and then continue into negligence, gross negligence, strict product liability for design defect and failure to warn, fraudulent misrepresentation, and public nuisance. The State alleges OpenAI marketed ChatGPT as safe and family-suitable while, it says, concealing internal safety warnings and shipping a product that was accessible to children without meaningful age verification. OpenAI has not been found liable on any of it, and the company disputes the State's characterizations.
Florida asks for civil penalties of up to $10,000 per willful violation, permanent injunctions, damages including treble and punitive damages, disgorgement and fees. Penalties at that rate multiply quickly against a user base measured in millions, which is why the State has publicly framed the exposure in the billions. None of that money would reach individual consumers — civil penalties under FDUTPA are payable to the State. OCA's full read of the complaint is here: Florida Sues OpenAI Over ChatGPT — Is It a Class Action?
Eleven days later, on June 12, 2026, a coalition of state attorneys general reported to number 42 served OpenAI with a broad investigative subpoena, spearheaded by New York Attorney General Letitia James. Reporting describes it as reaching advertising practices, engagement and retention design, consumer and health data handling, treatment of minors and older users, and model sycophancy. OpenAI has said it is engaging with the states. An investigative subpoena is not an accusation and creates no liability.
It does, however, follow a well-worn path. Multistate AG investigations of consumer-facing technology companies have historically ended in negotiated resolutions that combine a payment to the participating states with injunctive commitments. That is the mechanism by which a state-directed OpenAI payment would most plausibly arrive.
Track Three — Consumer Class Actions, and the Arbitration Problem
This is the track most readers assume will produce a check, and it is the one least likely to do so before 2028.Consumer class litigation against OpenAI has been thin and unstable. A proposed class action accusing OpenAI of routing ChatGPT queries to Meta and Google through embedded tracking code was voluntarily dismissed by the named plaintiff eight days after filing — OCA covered it in ChatGPT Lawsuit: Did OpenAI Share Chats With Meta and Google? Privacy theories against AI assistants remain live across the sector, but none has produced a certified consumer class against OpenAI.
The structural obstacle is contractual. OpenAI's consumer terms require final and binding individual arbitration, waive class actions, and waive jury trial, with a limited window to opt out after account creation or after an update to the arbitration terms. Most users never opt out. A class-action waiver of that kind, enforced, converts what would have been a single classwide case into thousands of individual arbitrations — which is why consumer class claims against companies with well-drafted arbitration clauses so often never reach certification, and why mass-arbitration campaigns have become the substitute.
Three things that clause does not do. It does not bind state attorneys general, who sue in their own sovereign capacity and are not parties to a user agreement. It does not cleanly bind minors in every jurisdiction, since contracts with minors are voidable in many states and courts have been skeptical of enforcing them in injury cases. And it does not reach non-users — a bystander harmed by another person's use of the product never agreed to anything.
The Copyright Cases Are Not Tort Claims — But They Set the Price
OpenAI's largest headline exposure is still copyright. Book-author cases against OpenAI and Microsoft were centralized as In re OpenAI, Inc., Copyright Infringement Litigation, MDL No. 3143, in the Southern District of New York, where the court denied dismissal of the core infringement claim and moved the cases into discovery. The New York Times v. Microsoft and OpenAI, filed in December 2023, is further along: summary-judgment briefing has run through 2026, and Judge Sidney Stein is expected to decide what, if anything, goes to trial.The direction of travel there turned in OpenAI's favor on September 1, 2026, when the U.S. Department of Justice filed a statement of interest backing the position that training a model on copyrighted text can be fair use, describing the training as highly transformative. A statement of interest does not bind the judge and does not decide the case. But it is the first time the federal government has taken a side on AI training and copyright, and it lowers the probability of a forced nine- or ten-figure copyright payment on the Anthropic model.
Those cases matter to this question in an indirect way. A company facing a plausible multibillion-dollar copyright judgment has a reason to conserve cash and fight everything; a company that has just watched its copyright risk get cheaper has more room to settle the injury cases that carry reputational and regulatory consequences it cannot litigate away.
The Comparators: What Similar Defendants Actually Did
Forecasting from one company's docket is guesswork. Forecasting from what comparable defendants did in comparable postures is less so. Three comparators are directly on point.• Character.AI and Google. The closest analogue. Garcia v. Character Technologies was filed in the Middle District of Florida in October 2024 over a 14-year-old's suicide after months of interaction with a companion chatbot. In January 2025 the court largely refused to dismiss on First Amendment grounds, letting the product-liability theory proceed. Filings unsealed on January 7, 2026 disclosed mediated settlements in five cases across Florida, Colorado, New York and Texas. Terms are confidential, with no admission of liability, and the court dismissed subject to a 90-day window to finalize. Elapsed time from first filing to settlement: roughly fifteen months. OCA's coverage: Character.AI and Google Move to Settle Teen Chatbot Harm Lawsuits.
• Meta in New Mexico. A state child-safety case against a platform that went the distance and produced a $375 million verdict for the State — evidence that a state AG claim on child-safety design can be tried and won, and evidence of the order of magnitude. See the New Mexico verdict against Meta.
• Anthropic. $1.5 billion, final approval July 20, 2026 — but copyright, resolved on statutory damages, and therefore not a guide to tort timing.
The Character.AI comparator is the one that carries the most weight, because the posture is nearly identical: a chatbot developer, a teen-harm wrongful-death claim, a surviving motion to dismiss, a well-funded corporate co-defendant, and a decision to settle confidentially rather than let a jury hear the internal documents. That is a settled pattern, not a one-off.
The Timeline Math
Getting from where the dockets sit today to money before December 31, 2027 requires specific things to happen in specific windows.For an individual settlement in JCCP 5431, the sequence is: leadership appointed (done, August 2026), coordinated discovery, bellwether selection, dispositive motions, then either a trial date or a mediation. Coordinated proceedings of this size typically reach a first bellwether trial roughly two to three years after coordination, which places 2028 to 2029 as the realistic trial window from a February 2026 coordination order. But settlements do not wait for trials. They cluster around the moment internal documents become discoverable and a trial date becomes real — and that moment, for JCCP 5431, lands squarely inside 2027.
For Florida, a ten-count complaint filed June 1, 2026 in state court faces motions to dismiss testing whether a chatbot is a "product" for strict-liability purposes and whether its outputs can be a "public nuisance." Those are genuinely unsettled questions, and whichever way the trial court rules, an appeal is likely. State consumer-protection enforcement actions against large technology defendants routinely take three to five years to resolve. A Florida payment before 2028 would require an unusually fast negotiated resolution.
For the multistate investigation, the subpoena issued June 12, 2026. Document production, follow-up demands and negotiation in investigations of this breadth generally run two years or more before any resolution is announced. Late 2027 is the earliest plausible landing point and is on the optimistic end.
For a certified consumer class, the sequence would be: a surviving complaint, defeat or waiver of the arbitration clause, a contested certification motion, and then either Rule 23(f) appellate review or settlement. Nothing in that sequence has started. Even on a favorable path, a claimable consumer fund before 2028 is not achievable from a standing start today.
What Would Make an Early Payment More Likely
• The IPO. OpenAI is reported to have filed a confidential S-1 with the SEC in 2026 with a listing targeted for late in the year. Public-company diligence is hostile to open-ended personal-injury exposure. Underwriters and auditors press issuers to quantify and, where possible, retire contingent liabilities before pricing. That is one of the most reliable settlement accelerants there is.• The insurance gap. Reporting in late 2025 indicated OpenAI held emerging-AI-risk cover of up to roughly $300 million, a figure one person familiar with the policy disputed as too high, and that the company had discussed a captive insurance vehicle and self-insurance using investor funds. Either way, the coverage described is small relative to the claimed exposure. A defendant paying from its own balance sheet, without an insurer's litigate-everything reflex, settles earlier than one whose carrier controls the defense.
• Discovery risk. The Florida complaint and the JCCP pleadings both allege that internal safety personnel raised objections that were overridden. If coordinated discovery surfaces documents supporting that account, the settlement value of every case in the inventory moves at once.
• The precedent next door. Character.AI and Google already chose confidential settlement over a jury trial in materially the same posture. Defense counsel notice that.
• Regulatory convergence. A 42-state investigation running alongside private injury litigation creates pressure to resolve both together, because a contested trial generates the factual record the AGs would otherwise have to build themselves.
What Would Push It Past 2028
• Causation. This is the hardest problem in every one of these cases. Proving that a chatbot conversation, rather than an underlying mental-health condition or any of a dozen other factors, caused a specific death or injury is a genuine evidentiary burden. It is the issue OpenAI has pressed from the start, and it is the reason these cases are not simply a matter of computing damages.• The "is it a product?" question. Strict product liability applies to products. Whether generative text output is a product or protected expression is unsettled. A defense win on that question in an appellate court would reshape the entire docket.
• First Amendment and Section 230 defenses. These have not succeeded broadly yet — the Character.AI court declined to dismiss on First Amendment grounds at the pleading stage — but the pleading stage is not the last word, and an appellate ruling could revive them.
• Arbitration. Enforced against adult users, it removes the consumer-class path almost entirely.
• A litigate-everything posture. A company that believes an adverse verdict would be existential for the industry may prefer to fight, appeal, and absorb the cost. OpenAI has litigated aggressively so far.
The Bottom Line — Three Questions, Three Answers
The single-sentence answer is: probably yes, but almost certainly not in the form most people picture.Will OpenAI pay individual tort claimants before 2028? More likely than not. The Character.AI precedent is directly on point, the JCCP inventory is growing, an IPO creates strong pressure to retire contingent liabilities, and confidential individual settlements require no class, no court approval of a fund and no admission. The most probable form is exactly what happened next door: a set of quiet, confidential resolutions of the earliest-filed cases, disclosed — if at all — only through a docket entry.
Will a state receive money from OpenAI before 2028? Possible, but less likely than not. Florida's case is barely past the pleading stage on legal theories no appellate court has resolved. The multistate investigation is at the document-production stage. Both tracks can plausibly produce a payment, but 2028 to 2030 is the more realistic window for either.
Will there be a claimable OpenAI consumer class settlement — a fund, an administrator, a claim form — before 2028? Unlikely. No class is certified, the arbitration clause with its class waiver stands between users and certification, and tort causation is individualized in a way that resists classwide treatment. Nothing currently on file is positioned to produce a consumer claim form inside this window.
The distinction between the first and third answers is the whole point. "OpenAI paid" and "you can file a claim" are different events, and the first is much more likely to happen without the second.
What This Means for Consumers Right Now
There is nothing to file. No OpenAI or ChatGPT settlement exists, no fund has been created, no administrator has been appointed and no deadline is running. Any site inviting you to submit a ChatGPT settlement claim today is not describing an approved settlement, and OCA's guide to why AI chatbots get class action deadlines wrong covers how these phantom deadlines propagate.A family that believes an AI chatbot contributed to a real injury is in a different position from a general ChatGPT user. That is an individual personal-injury claim with its own state-law filing deadline, not a class membership, and the deadline runs whether or not any coordinated proceeding exists. Individual legal advice on that is a matter for a licensed attorney in the relevant state.
The concrete things worth watching over the next fifteen months: the September 23, 2026 case management conference in JCCP 5431 and whatever bellwether schedule follows it; any ruling on OpenAI's response to the Florida complaint, particularly on whether a chatbot is a product; Judge Stein's summary-judgment ruling in the Times case; and any announcement from the multistate attorney general group. Those four are where the answer to this question will actually be written.
Free settlement alerts
Get notified when new class actions open to claims
Join thousands of readers who get the latest class action settlements you may qualify for — delivered straight to your inbox.
Frequently Asked Questions
Has OpenAI ever paid damages to consumers?
No. As of September 17, 2026, there is no public record of OpenAI paying damages on a personal-injury, product-liability or consumer-protection claim, and no court has entered a damages judgment against it on such a claim. The company has settled and licensed its way out of some publisher disputes, but those are contract and copyright matters, not tort claims.
Is there a ChatGPT class action settlement I can file a claim for?
No. There is no OpenAI or ChatGPT settlement fund, no claim form and no deadline. No consumer class has been certified against OpenAI in a tort case. Any website inviting you to file a ChatGPT settlement claim today is not describing an approved settlement.
What is JCCP 5431?
JCCP No. 5431, captioned In re ChatGPT Product Liability Cases, is a California Judicial Council coordinated proceeding created by an order in San Francisco in February 2026. It gathers individual product-liability and wrongful-death suits against OpenAI before one judge in San Francisco Superior Court for pretrial purposes. It is not a class action, so there is no class to join and nothing to claim.
Does the Florida attorney general's lawsuit mean Floridians get money?
No. Florida's June 2026 suit is a state enforcement action. The civil penalties it seeks are payable to the State, not to individual consumers, and the injunctions it asks for would change how OpenAI operates rather than create a fund. A state enforcement action does not produce a consumer claim form.
Can ChatGPT users bring a class action, or does arbitration block it?
OpenAI's consumer terms require individual arbitration and waive class actions and jury trials, with a limited window to opt out after account creation or after an update to the arbitration terms. That clause is the single biggest structural obstacle to a claimable consumer class fund. It does not bar state attorneys general, who are not bound by a private contract, and courts sometimes decline to enforce arbitration clauses against minors or on unconscionability grounds.
Why does the Anthropic settlement not answer this question?
Anthropic's $1.5 billion settlement, granted final approval on July 20, 2026, resolved copyright claims brought by authors over books used in training. Copyright carries statutory damages that can be calculated per work without proving anyone was hurt. A tort claim requires proof that a specific person suffered a specific injury the product caused, which is a far harder and slower case to value.
Sources
• Docket, In re: OpenAI, Inc., Copyright Infringement Litigation, No. 1:25-md-03143 (S.D.N.Y.), via CourtListener• Coordination order, In re ChatGPT Product Liability Cases, JCCP No. 5431 (Cal. Super. Ct., San Francisco)
• Press release, Florida Attorney General, announcing the state-led lawsuit against OpenAI and its CEO (June 1, 2026)
• Complaint, Office of the Attorney General, State of Florida, Department of Legal Affairs v. OpenAI Global, LLC, et al., Circuit Court of the Tenth Judicial Circuit, Highlands County, Florida (filed June 1, 2026)
• Florida Deceptive and Unfair Trade Practices Act, Section 501.201 et seq., Florida Statutes
• NPR, "Florida sues OpenAI and Sam Altman over alleged safety lapses" (June 1, 2026)
• CNBC, on the multistate attorney general subpoena and OpenAI's response (June 12, 2026)
• CNN Business, "Character.AI and Google agree to settle lawsuits over teen mental health harms and suicides" (January 7, 2026)
• Bloomberg Law, "Character.AI, Google Agree to Settle Teen Chatbot Harm Lawsuits"
• Docket, Garcia v. Character Technologies, Inc., No. 6:24-cv-01903 (M.D. Fla.), via CourtListener
• Axios, on the posture of The New York Times v. Microsoft and OpenAI (September 8, 2026)
• OpenAI Terms of Use — arbitration agreement, class-action waiver and opt-out window
• Financial Times reporting (October 2025) on AI-company insurance capacity, OpenAI's emerging-risk coverage, and captive/self-insurance discussions
About This Page
This is an analysis of pending litigation, not a prediction anyone should rely on for a legal or financial decision. Every case described here is unresolved: the allegations in Raine v. OpenAI, in the other cases coordinated as JCCP No. 5431, and in the State of Florida's complaint are the plaintiffs' and the State's claims, they are unproven, and OpenAI and Sam Altman are entitled to respond and defend in court. No court has found OpenAI liable on any of them. OpenClassActions.com is a consumer news and information site — not a law firm, not a party to any of these cases, and not counsel to anyone. This page is general information, not legal advice, and nothing on it is tax advice. There is no OpenAI or ChatGPT settlement, fund, administrator or claim form, and no deadline is running for consumers. If you or someone you know is struggling, help is available in the U.S. by calling or texting 988, the Suicide and Crisis Lifeline.
For more class actions keep scrolling below.
