Attorneys general in eight states filed separate lawsuits against OpenAI on October 14, each in its own state court, arguing that the company’s chatbot is harming children.
Iowa, Kansas, Nebraska, West Virginia, Arkansas, Arizona, Idaho and Montana brought the cases under state consumer protection and fraud statutes. The complaints describe a system that is designed to defer to users, that collects data from minors and that offers dangerous medical advice. The relief sought includes the disgorgement of profits earned during the period at issue and stronger protections for children.
Brenna Bird, Iowa’s attorney general, said the state intends to hold the company accountable for conduct she described as deliberate, dangerous, reckless and unlawful. A company spokesman, Drew Pusateri, said the product includes age-based protections that were built in.
Filing in eight separate courts rather than jointly is a choice with practical consequences. A multistate action would rest on shared legal theories and a common record. Eight cases let each state tailor its claims to its own consumer protection statute, its own evidentiary rules and its own appetite for a trial. It also multiplies the discovery burden on the defendant, which is a consideration in any negotiation about how the cases end.
The central allegation concerns design rather than a single harmful output. A persuasive system built to be agreeable will agree with a user who expresses a dangerous intent, which is the tension at the center of how conversational products are tuned. Engineers optimize for helpfulness because unhelpful answers drive users away, and the same optimization produces responses that states describe as dangerous.
The data claims belong to a second body of law. Children’s privacy rules in the United States restrict the collection of personal information from users under 13, and enforcement has historically run through a federal regulator. States have argued in other cases that their own consumer protection statutes reach data practices that federal rules also cover, and courts have allowed those claims to proceed when the conduct alleged is deceptive rather than merely unlawful.
Medical advice is the third theory and the easiest for a jury to understand. A system that answers a question about symptoms with a confident suggestion can delay treatment, and the documentation of those exchanges sits in conversation logs that the plaintiff’s lawyers will seek. Whether the company warned users away from treating the output as clinical advice becomes a question of what the interface said and how prominently it said it.
The stakes for the company extend past damages. State attorneys general can seek injunctive relief that changes a product’s design, which would make an engineering decision for the defendant. Agreeing to age verification, to limits on certain conversations for younger users or to reporting requirements would set terms that other jurisdictions and other developers would then be measured against.
The federal government is moving in parallel. The Federal Trade Commission has demanded records from the largest developers about safety incidents and protective measures, and the agency has said it plans to compel testimony from executives. Federal and state actions pursuing the same conduct raise questions about coordination that defendants typically raise early in litigation.
Other developers are watching for the same reason. The claims here apply to a category of product rather than to one company’s choices, and a ruling that a conversational interface is deceptive when it fails to warn about a foreseeable harm would apply to every competitor. That prospect tends to produce industry-wide changes in disclaimers and default settings.
Analysts who follow consumer litigation said the most likely outcomes are settlements with injunctive terms, because states rarely take these cases to verdict and companies rarely prefer a public trial about children. The shape of a settlement would become the de facto standard, adopted by competitors without a court order.
What the complaints leave open is causation. Proving that a specific conversation produced a specific harm requires linking a product’s output to conduct, and the intervening choices a user makes are difficult to establish in court. States that have brought similar cases against other consumer products have often relied on the design of the product itself rather than on an individual outcome, which is the path these filings appear to follow.
The cases also test how state law applies to a service that is not sold as a product. A user under 18 typically reaches the chatbot through a free account, so there is no purchase and no price against which to measure a loss. Consumer protection statutes in several states nonetheless cover unfair practices as well as deceptive ones, and the remedies available under those provisions include civil penalties per violation, a figure that scales with the number of accounts.
The company’s response focused on what the product already does. A spokesman said age-based protections are built into the system, which is the argument a defendant makes when the claim concerns design choices rather than a single failure: the safeguards exist, they are documented, and the plaintiffs disagree about whether they are sufficient.
The states that filed are also sending a message to their own legislatures. Several have considered bills on chatbot regulation for minors, and a lawsuit filed while a bill is pending gives the attorney general a record and a remedy without waiting for a vote. The filings can be read as an argument that existing law already covers the conduct, which is the position a regulator takes when the legislative calendar is slow.
