Loading page…
The cases target safety claims, chatbot design and children’s data. Arkansas is also seeking restrictions on advanced AI testing until safeguards are in place.
Listen to this story
Arkansas’s requested remedies would reach beyond ChatGPT’s responses to minors: they include barring collection of minors’ data, restitution and civil penalties, and pausing advanced AI testing and recursive-self-improvement models until safeguards are adequate. Arizona and Arkansas allege deceptive safety claims and child-related harms; Arkansas also invokes children’s privacy law. The suits remain allegations, not findings. The number of state plaintiffs is unclear: Arkansas lists nine filings including its own, while The Guardian reported eight.
The lawsuits were filed on October 14, 2026, and seek court-ordered changes to OpenAI’s practices for young users.
Arizona alleges ChatGPT’s agreeableness and simulated compassion can encourage compulsive use, personal-data sharing, and reluctance to seek help from others.
Arkansas alleges ChatGPT provided self-harm instructions and that OpenAI collected and retained children’s personal information without required notice or consent.
US state attorneys general sued OpenAI on October 14, 2026, alleging ChatGPT endangered children and seeking changes to how it operates for young users. The cases challenge more than individual chatbot answers: official announcements from Arizona and Arkansas accuse OpenAI of misleading consumers about safety, while the states also seek profits generated from young users’ activity.
The Guardian reported eight states had sued, including Kansas, Arizona and Montana. Arkansas’s official announcement gives a larger count: it names its own filing and similar lawsuits filed that day by Iowa, Kansas, Nebraska, West Virginia, Arizona, Idaho, Montana and Utah. That list totals nine states, a discrepancy with the early news coverage.
Arizona Attorney General Kris Mayes filed her case in Maricopa County Superior Court. Her office’s announcement alleges violations of the Arizona Consumer Fraud Act through ChatGPT’s design and marketing, including concealed features that make the chatbot unusually agreeable and simulate human compassion to keep users engaged.
The alleged mechanism is a trained tendency to agree with users that, Arizona says, overwhelms existing safety guardrails. The complaint accuses OpenAI of engineering compulsive use among minors and encouraging them to share personal information with ChatGPT while sometimes discouraging disclosure to parents, relatives, authorities or friends.
Arizona’s allegations extend beyond self-harm. Its announcement lists diminished critical thinking, unwarranted paranoia or delusional thinking, dangerous unlicensed medical guidance, and unhealthy behaviors and relationships. It says some outputs were interpreted by users as encouraging self-harm or violence toward others—a narrower allegation than treating every such output as an explicit instruction.
The state maintains that these risks were reasonably foreseeable to OpenAI but were not communicated to Arizona consumers. Its case connects the alleged harms to the public’s reliance on representations that ChatGPT was safe and appropriate for young people.
But one thing that has not changed and will not change is Arkansans’ right to receive truthful information about the products and services they use.
Arkansas Attorney General Tim Griffin
Arkansas Attorney General Tim Griffin’s announcement alleges that consumers, including minors, obtained self-harm instructions from ChatGPT. It also describes users bypassing safety measures to generate responses OpenAI said its safeguards were designed to prevent. Griffin argues that OpenAI knew of significant risks and limitations but hid that knowledge from consumers.
The complaint invokes the Arkansas Deceptive Trade Practices Act and the Children’s and Teens’ Online Privacy Protection Act. The privacy allegations include knowingly collecting children’s and teens’ personal information, keeping it longer than reasonably necessary, and failing to provide notice and obtain consent. Other claims include negligence, defective design, failure to warn, unjust enrichment and public nuisance. These are the state’s allegations, not established findings.
The requested remedies differ in scope. Alongside the operational changes and surrender of profits described in the Guardian’s coverage, Arkansas sets out specific demands covering data collection, consumer warnings, compensation and the development of more advanced systems.
Story updates
OpenAI faces a demand for court-ordered changes to ChatGPT’s safeguards, alongside financial penalties, in a West Virginia consumer-protection lawsuit. Attorney General JB McCuskey filed the action on October 14, 2026, alleging the company misled families about safety and failed to adequately protect children and other vulnerable users.
The civil complaint was filed in Mason County Circuit Court and names eight OpenAI-related corporate entities as defendants. The allegations have not been established in court. West Virginia is asking a judge to determine whether the company violated state law and to order remedies if its claims are proven.
The state’s case reaches beyond whether ChatGPT gives accurate answers. It challenges how the chatbot handles conversations about suicide, self-harm, violence and other sensitive subjects. It also questions whether the product’s conversational design can encourage emotional dependence, making the way it interacts with users part of the alleged consumer-protection problem.
In particular, the complaint targets ChatGPT’s tendency to agree with users and simulate human empathy. West Virginia argues those characteristics can encourage excessive use or emotional dependence among minors and other vulnerable people. The attorney general’s announcement accuses OpenAI of presenting ChatGPT as a product that puts safety first while exposing vulnerable users to danger.
According to Charles Young’s October 19 report in WV News, the filing links the technology to alleged psychological harm, delusions, medical misinformation and financial losses. It also attributes at least 24 deaths worldwide to incidents involving the technology. That figure is the state’s allegation, not a judicial finding that ChatGPT caused those deaths.
The requested relief would reach the service itself, not just OpenAI’s finances. West Virginia wants a court-supervised plan addressing youth digital safety and crisis response. Its proposed protections cover several parts of the user experience:
Stronger age verification, addressing the state’s challenge to how the service identifies young users.
Default parental controls, making those protections a starting condition rather than simply seeking additional disclosures about risk.
Protocols for responding to users in crisis, alongside the state’s challenge to conversations involving self-harm and other sensitive subjects.
Independent monitoring of OpenAI’s compliance, adding outside oversight to the safeguards the state wants the court to require.
The state also seeks deletion of personal data belonging to West Virginia minors that it alleges was unlawfully collected. That request concerns information already held by the company, distinct from the proposed changes to age checks, parental controls and future conversations.
West Virginia is pursuing the case under its Consumer Credit and Protection Act and other legal theories. The complaint alleges deceptive and unfair trade practices, public nuisance and unjust enrichment. It argues that OpenAI prioritized commercial growth over safety, with safety reviews receiving less attention as the company pursued revenue.
Beyond civil penalties, the state seeks restitution, the return of profits allegedly obtained through unlawful conduct, and reimbursement of legal and investigative costs. The requested relief therefore combines financial recovery with operational restrictions: the judge is being asked both to address alleged past violations and to require protections governing the product.
wv.govLoading discussion...
Join the conversation
Explain where you would draw the line between protecting users and restricting research.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.