Oregon sues OpenAI for $2.035 billion over ChatGPT child safety

Oregon Attorney General Dan Rayfield sued OpenAI on November 17, 2026, accusing the company of misleading parents about the safety of ChatGPT for children and teenagers and asking the Multnomah County Circuit Court for damages of not less than $2.035 billion. The complaint, filed on behalf of the State of Oregon, names six OpenAI entities as defendants: OpenAI Global, LLC; the OpenAI Foundation (formerly OpenAI, Inc.); OpenAI OpCo, LLC; OpenAI Group PBC; OpenAI Holdings, LLC; and OpenAI, LLC. All six are described as Delaware entities with their principal place of business in San Francisco.
The state brings five claims. Two rest on Oregon's Unlawful Trade Practices Act, which allows the attorney general to seek injunctions, civil penalties for each willful violation and restitution. The other three, negligence, unjust enrichment and public nuisance, are brought in the state's parens patriae capacity, meaning Oregon sues on behalf of its residents, in this case its minor ChatGPT users. The $2.035 billion figure appears in the negligence and public nuisance counts as a floor, "not less than" that sum, to be proved at trial. The complaint does not explain how the amount was calculated and does not itemize penalties, restitution or disgorgement.
Rayfield announced the case at a press conference as the centerpiece of a new "Keeping Kids Safe" initiative of the Oregon Department of Justice. "OpenAI knows that ChatGPT is harming children, and has failed to meaningfully fix their dangerous platform," he said in the state's announcement. The filing is signed by Rayfield, two Oregon assistant attorneys general and four lawyers from the firm Labaton Keller Sucharow, whose admission to appear in the case is pending. The state has demanded a jury trial.
The lawsuit lands one day after California filed its own case against OpenAI over AI agents, and it is the latest in a run of state actions targeting how ChatGPT is built and marketed. According to the complaint, ChatGPT has "hundreds of thousands" of users in Oregon, and OpenAI's own data ranked the state 25th in U.S. usage in the third quarter of 2026.
··········
THE FIVE CLAIMS IN OREGON'S CHATGPT COMPLAINT.
The core of the case is consumer-protection law. In its first claim, the state alleges that OpenAI misrepresented the characteristics, uses, benefits and qualities of ChatGPT under ORS 646.608(1)(e), and failed to disclose a known material defect under ORS 646.608(1)(t). The complaint says OpenAI acted willfully, because it knew or should have known that its safety representations were false or misleading. The state also says it gave OpenAI the pre-suit notice the statute requires and did not receive a satisfactory assurance of voluntary compliance, the formal settlement commitment that would have allowed the matter to close without litigation.
The second claim relies on the act's unconscionability provision, ORS 646.607(1). Here the state describes three tactics: exploiting users' mental and physical vulnerabilities, exploiting children's developmental immaturity without adequate safeguards, and combining design choices that the complaint lists as anthropomorphism, sycophancy, memory, personalization and inadequate age verification.
The remaining three claims move from statute to common law. The negligence count alleges that OpenAI owed minor Oregon residents a duty of care and breached it. The unjust enrichment count alleges that OpenAI kept benefits it obtained from minor users, including their data and their payments, and asks the court to order those proceeds disgorged without naming a figure. The public nuisance count cites Article I, sections 1 and 8, of the Oregon Constitution and repeats the $2.035 billion floor.
........
Claim | Legal basis | What Oregon asks for |
1. Misrepresentation and nondisclosure | UTPA, ORS 646.608(1)(e) and (1)(t) | Injunction, civil penalties per willful violation, restitution |
2. Unconscionable tactics | UTPA, ORS 646.607(1) | Injunction, civil penalties per willful violation, restitution |
3. Negligence | Common law, parens patriae | Damages of not less than $2.035 billion |
4. Unjust enrichment | Common law, parens patriae | Disgorgement of proceeds from minor users |
5. Public nuisance | Common law, Oregon Constitution Art. I | Damages of not less than $2.035 billion |
........
All claims are brought against the six defendants jointly and severally, so the state can in principle recover any judgment from any of them. The structure reflects OpenAI's corporate reorganization: the complaint notes that the restructuring finalized in October 2025 left the OpenAI Foundation holding about 26% of the equity of the for-profit group, and Oregon has chosen to sue the nonprofit parent alongside the operating companies.
··········
WHAT OREGON WANTS A COURT TO FORCE OPENAI TO CHANGE IN CHATGPT.
The money figure drew the headlines, but the prayer for relief is where the case would touch the product. Oregon asks the court for a permanent injunction under ORS 646.632 and 646.636 that would bar OpenAI from misrepresenting ChatGPT's safety, its suitability for different ages and the effectiveness of its parental controls, and from failing to disclose known risks to children.
The state then asks for affirmative obligations. The complaint requests age-assurance measures, safeguards against harmful content, accurate disclosures of risks including emotional dependence, self-harm, disordered eating and violence, and parental controls and safety notifications that work as advertised. Beyond the injunction, Oregon seeks compensatory and statutory damages for residents, civil penalties under ORS 646.642(3) for each willful violation, pre- and post-judgment interest, attorneys' fees and costs.
If a court granted the injunctive part of that request, OpenAI would face Oregon-specific obligations on age checks and parental tools. Because ChatGPT is a single national service, companies in that position often apply such changes more broadly than the one state that obtained them, though nothing in the filing requires that.
··········
THE CHATGPT DESIGN FEATURES AND STUDIES THE COMPLAINT CITES.
Much of the filing argues that specific design decisions made ChatGPT riskier for young users. The complaint points to anthropomorphic design, meaning human-like responses, expressed feelings and cues of empathy, which the state says can create emotional dependence. It cites sycophancy, the tendency of a model to agree with and flatter the user, including OpenAI's April 2025 rollback of a GPT-4o update, and a March 2026 study that found ChatGPT endorsed users' actions about 40% more often than crowdsourced human responses. It also cites memory, which the complaint says has been on by default since 2024 and was expanded later.
To show how widely children use these tools, the state relies on outside research. The figures below are the complaint's own citations, not independent findings by Data Studios.
........
Figure cited in the Oregon complaint | Value | Study cited |
U.S. teens aged 13-17 who have used AI chatbots | 72% (52% at least monthly) | Survey of 1,060 teens |
Teens using chatbots for social or relationship purposes | 33% | Same survey |
Children aged 10-12 using generative AI | 20.5% (9.4% at ages 8-9) | JAMA Network Open, Feb. 2026, 6,488 children |
Share of those young users on ChatGPT | 78.8% | Same study |
Children aged 9-12 who had used ChatGPT | 39% | Internet Matters, July 2025 |
Oregon's rank in U.S. ChatGPT usage | 25th, Q3 2026 | OpenAI data |
........
On safety testing, the complaint cites an August 2025 report by the Center for Countering Digital Hate that rated more than half of the ChatGPT responses it tested as dangerous, a separate study that found 53% of 1,200 responses harmful, and testing showing that users could get past refusals by claiming a harmless purpose such as preparing a presentation. The state also cites research on learning: a study of 26,811 Chinese students that estimated a 20% drop in closed-book exam scores within six months of adopting generative AI, and a study of 63 students aged 14 and 15 who asked follow-up questions after inadequate answers only about 8% of the time.
The complaint also refers to several deaths of minors and young adults and to two shootings, drawing on other lawsuits and public reports rather than on cases in Oregon. Those incidents are allegations in litigation and have not been tested in this court.
··········
[[ADPLUS_300x250_1]]
CHATGPT FOR TEENS, AGE PREDICTION AND PARENTAL CONTROLS UNDER SCRUTINY.
A large part of the case targets the protections OpenAI introduced for younger users. According to the complaint, before August 2026 ChatGPT did not verify parental consent when users aged 13 to 17 signed up, and children under 13 could create accounts. OpenAI launched age prediction, a system that estimates whether an account belongs to a minor, in August 2026. The complaint cites Common Sense Media, which reported that it could not get the system to classify an adult account as a minor even after thousands of attempts.
The state gives particular attention to ChatGPT for Teens, which OpenAI launched on August 18, 2026. The complaint argues that the product's protections depend on a parent linking their account, and that this linking is optional, so a teenager can use the service with no parent connected. It cites Common Sense Media's October 7, 2026 assessment that the product poses an "unacceptable risk," and testing in which 25% of responses to more than 200 crisis-type prompts included no hotline, professional or medical resource. Testers also reported that parental alerts were not sent during some high-risk conversations and that alerts which did arrive were delayed, and that the product's study mode could be bypassed by removing a prefix from the prompt.
OpenAI has added several teen-specific limits during 2026, some of which we covered when the company restricted weapons information for teen ChatGPT users and added chat-ending for violent conversations. Oregon's argument is that these measures, as shipped, do not match the way OpenAI has described them to parents. The press release and the complaint contain no response from OpenAI, and the company's answer will come in its court filings.
··········
WHAT THE OREGON LAWSUIT MEANS FOR CHATGPT USERS AND PARENTS.
For people using ChatGPT today, nothing changes as a direct result of the filing. A complaint is a set of allegations, the court has made no findings, and no injunction is in force. ChatGPT remains available in Oregon on the same plans and terms as elsewhere in the United States, including the free tier that most teenagers are likely to use, which we describe in our guide to what the ChatGPT free plan includes.
The Oregon Department of Justice did publish practical advice for parents alongside the lawsuit. It recommends asking children whether they use ChatGPT or other AI chatbots and what they talk about, watching for gamified features designed to encourage repeat use, and not assuming that parental controls or age verification block all inappropriate content. For families that want ChatGPT's teen protections, the complaint's own account suggests the practical step: link the parent and teen accounts, since several of the safeguards it describes only operate when that link exists. Anyone in distress in the United States can call or text 988 to reach the Suicide and Crisis Lifeline, or text HOME to 741741 for the Crisis Text Line.
For companies building on OpenAI's models or deploying ChatGPT in schools, the case is a signal about where state regulators are focusing: age assurance, the gap between marketing claims and product behavior, and default-on features such as memory. Oregon's request that disclosures cover emotional dependence explicitly is a theme that also runs through other state actions, including the separate case California filed against OpenAI over rogue AI agents a day earlier.
··········
THE KEEPING KIDS SAFE INITIATIVE BEHIND THE OPENAI CASE.
Rayfield presented the OpenAI suit as one part of a wider program announced with Multnomah County District Attorney Nathan Vasquez, Representative April Dobson, Oregon PTA president-elect David Jones and Oregon Education Association president Enrique Farrera. The department cited a $125 million settlement with Meta that limits daily use of Instagram and Facebook for young users, blocks overnight access and stops push notifications during school hours, and said it is working to apply similar standards to other large social media companies.
The initiative also includes an investigation of Discord opened earlier in 2026, an expansion of the state's Internet Crimes Against Children task force from about five full-time staff to 20 over two years, and a partnership with local law enforcement against organized crime that the department says has closed seven businesses across five counties. Rayfield said more actions against companies that harm children are expected in the coming weeks and months, which places AI chatbots in the same enforcement category Oregon has been applying to social networks.
··········
OREGON V. OPENAI: CONFIRMED FACTS AND OPEN QUESTIONS.
The verified record is the filing itself: a five-count complaint dated November 17, 2026 in Multnomah County Circuit Court against six OpenAI entities, two counts under the Unlawful Trade Practices Act, three common-law counts brought on behalf of minor residents, a damages floor of $2.035 billion in two of them, and a detailed request for court-ordered age assurance, risk disclosures and working parental controls.
Several points remain open. The complaint does not show how the $2.035 billion figure was derived, how many willful violations the state will try to prove, or what civil penalties it will seek per violation. The studies it relies on come from advocacy groups, academic researchers and journalists, and their methods will be contested once OpenAI responds. The incidents it describes are drawn from other litigation and reports. The decisive questions for ChatGPT's design are whether a court accepts that OpenAI's public safety statements were misleading under Oregon law, and whether any injunction ends up requiring mandatory age assurance and parental linking that OpenAI has so far left optional.
··········
FOLLOW US FOR MORE.
·····
DATA STUDIOS
·····
datastudios.org




