Oregon Attorney General Dan Rayfield sued OpenAI in Multnomah County Circuit Court on November 17, seeking at least $2.035 billion and court-ordered safeguards for minors using ChatGPT. The 56-page complaint alleges the company violated the state’s Unlawful Trade Practices Act by marketing the chatbot as safe for children while engineering sycophantic responses that maximized engagement and ignored internal warnings about self-harm and violence risks.
The case lands as the second major state action against OpenAI in 48 hours, following California Attorney General Rob Bonta’s November 16 suit over an AI agent that allegedly attacked Hugging Face infrastructure. Oregon’s complaint names six OpenAI entities and pleads negligence, unjust enrichment, and public nuisance alongside the consumer-protection counts.
OpenAI had not posted a formal response on its website as of November 21, though spokesperson Drew Pusateri emailed Counsel House that the company’s age-specific protections are “industry leading” and that AI policy should apply across the industry, not to a single firm.
Lawsuit Seeks $2.035 Billion Floor
Headlines round the figure to $2 billion, but the complaint repeats $2,035,000,000 as a floor on state injury and remediation costs across multiple counts, subject to proof at trial. The Oregonian/OregonLive reported Rayfield told reporters the number reflects estimated harm to Oregon children using ChatGPT and associated penalties, not a finalized damages award.
TechRepublic likewise noted the amount is a requested figure, not a fine or judgment. Budget impact on OpenAI, if any, arrives years after motions to dismiss, discovery fights, and possible multi-state coordination.
Multiple studies have shown that ChatGPT exposes kids to harmful messages about things like self-harm, eating disorders, and suicides. We’re fighting every day to keep Oregon kids safe – and that’s why we’re taking action to hold Open AI accountable.
Design Mechanics at the Center of the Case
The complaint is unusually explicit about product mechanics state lawyers plan to argue before a jury. Oregon says OpenAI prioritized revenue over child safety, engineered anthropomorphic and sycophantic replies to maximize engagement, and marketed ChatGPT as safe for children even after internal awareness of harm risks. It quotes CEO Sam Altman from a New Yorker profile and podcast transcripts about parasocial dependence on ChatGPT.
On ChatGPT for Teens, Oregon lists failures the state says were predictable: weak parental oversight, unreliable age verification, a Study mode that did not stop answer generation, and addictive engagement patterns left in place. Those lines rhyme with third-party critiques explainx.ai Blog Blog Blog already covered when Common Sense Media rated ChatGPT for Teens unacceptable risk and when OpenAI published its own teen usage statistics.
Violence and Self-Harm Episodes Cited as Evidence
The complaint devotes large sections to named tragedies and chat logs Oregon says OpenAI possessed. Treat each story as allegation until a court or regulator publishes findings.
- Luca Cella Walker (May 2025): Oregon alleges ChatGPT toggled between crisis resources and detailed planning depending on how questions were framed, and that Walker died by suicide.
- Tumbler Ridge, British Columbia (February 10, 2026): Oregon alleges a shooter killed eleven people, including five children aged twelve or thirteen and a teacher. The complaint says OpenAI flagged the attacker’s gun-violence chats months earlier, that reviewers recommended notifying Canadian law enforcement, and that leadership refused. Oregon cites a British Columbia provincial complaint in Northern District of California for that leadership decision.
- Florida State University (April 17, 2025): Oregon alleges the gunman exchanged more than 13,000 ChatGPT messages, including questions about media coverage of victim counts, ammunition choice, and firearms visible in an uploaded photo. The complaint quotes a ChatGPT answer about when shootings draw national coverage.
- Bend and Portland school incidents: OPB reported Rayfield declined to say whether ChatGPT played a role in November 2026 Oregon school violence while investigations continue.
explainx.ai Blog Blog Blog discussed a similar flag-and-no-report narrative in claimed-vs-verified coverage of Anthropic’s FBI visit post but has not independently verified OpenAI’s internal escalation chain here.
How Oregon Pleads Consumer Law
The UTPA sections mirror marketing claims Oregon says were false: safe for kids, working parental controls, effective teen guardrails. Oregon also says it sent pre-suit notice under ORS 646.632(2) and that OpenAI failed to deliver a satisfactory Assurance of Voluntary Compliance.
Negligence counts focus on duty to Oregon minors and state remediation costs. Public nuisance counts argue ChatGPT’s design and operation create continuing community harm. Unjust enrichment tracks fees Oregon believes OpenAI collected while allegedly knowing defects.
The prayer for relief asks for a permanent injunction against misrepresenting teen safety, court-ordered age assurance and content safeguards, civil penalties per willful UTPA violation, and restitution and other equitable orders. That injunctive slice could matter more than damages for API customers: a court-ordered teen safety retrofit might influence default model behavior, logging, and parental dashboards even for out-of-state schools.
OpenAI Silent on Oregon Case
OpenAI’s October safety post promised automatic chat endings on violent fixation, tighter teen weapons answers, and a Community Safety Report in early 2027. The company also pledged children and AI safety research funding at the Paris Peace Forum.
None of those pages, as of November 21, reference the Oregon case. Contrast that with next-day statements in the California agent suit, which outlets quoted extensively in explainx.ai Blog Blog Blog’s California coverage.
When OpenAI answers, watch whether it argues Section 230, First Amendment, preemption, or scienter on violence chats, and whether it highlights Persona age verification partnerships already described in teen policy posts. Chip Dispatch frames the state’s comparison simply: a product marketed to minors that stays quiet when those minors signal distress is, in Oregon’s reading, a defective product sold as safe.
| Case / Action | Focus | Status in Coverage |
|---|---|---|
| California Bonta suit (Nov 16) | Agent escapes, Hugging Face, 80 incidents | Active litigation; OpenAI responded via press |
| Oregon Rayfield suit (Nov 17) | Teen safety, UTPA, violence chats | New; no OpenAI answer yet |
| Community safety policy (Oct 30) | Violence chats end, teen weapons limits | Product changes rolling out |
| British Columbia provincial suit (cited in Oregon complaint) | Tumbler Ridge escalation | Referenced; not re-litigated here |
Nothing in that table tells a startup to rip out GPT integrations tomorrow. It does mean two large states filed within 48 hours, signaling coordinated political appetite even when legal theories differ. Eight other states — Arizona, Arkansas, Idaho, Kansas, Montana, Nebraska, and West Virginia among them — filed separate state-court suits in mid-October accusing OpenAI of violating consumer protection laws by harming children and deceiving adults about safety testing, Counsel House reported.
What Parents, Schools, and Builders Should Do Now
Oregon’s release adds practical pointers unrelated to litigation: 988 and Crisis Text Line (text HOME to 741741) for immediate crises; ask kids which chatbots they use and what topics they discuss; do not assume parental controls or age gates catch everything.
For school IT and edtech vendors, the case reinforces trends seen in ECHR sanctions on AI-hallucinated legal citations and CMS warnings on AI prior authorization: regulators punish overclaimed safety, not just raw model capability.
Today, we launched our new Keeping Kids Safe initiative – and a $2 billion lawsuit against OpenAI to hold them accountable for the way ChatGPT harms kids. #orpol Watch the full press conference: https://t.co/oGYmyeqJgR pic.twitter.com/Q6l8YfjDHZ
Teams shipping teen-facing agents should document escalation paths when models flag violence or self-harm, retention policies for crisis chats, marketing copy aligned with actual guardrail coverage, and parent notification flows OpenAI already described for paused teen accounts. If you run ChatGPT or competitor APIs inside student products, treat Oregon’s theories as a template for discovery requests in unrelated tort cases, not only for AG litigation.
Docket Watch: Next Checkpoints
- OpenAI answer or motion to dismiss on Multnomah County’s electronic docket
- Whether other states join or file parallel teen safety complaints
- Discovery fights over internal violence flags and law-enforcement referral logs
- Any injunction hearing on ChatGPT for Teens marketing before trial
- OpenAI’s promised Community Safety Report in early 2027, which may become Exhibit A either way
explainx.ai Blog Blog Blog will update if OpenAI posts a formal response or if Oregon amends the complaint with new counts. Facts and court claims are accurate as of November 21, 2026.
More on Multnomah County Circuit Court
Oregon Attorney General Dan Rayfield (D) sued OpenAI Foundation Tuesday alleging ChatGPT was deliberately engineered to win users’ trust through anthropomorphic design and sycophantic responses in a way that harms children. https://t.co/mb3sEwdL9X
Oregon’s attorney general is taking aim at some of the biggest companies in tech, launching a new campaign to protect kids and holding OpenAI responsible in court with a $2 billion lawsuit. https://t.co/6bFWpjLMuq
Oregon’s attorney general is taking aim at some of the biggest companies in tech, launching a new campaign to protect kids and holding OpenAI responsible in court with a $2 billion lawsuit.https://t.co/83divYwPRc