Know what is on your site before a plaintiff’s firm does.Free website scanScan Your Site
Log in Sign up Book a demo
Home› News› Oregon Sues OpenAI Over ChatGPT’s Risks to Kids…
NEWS

Oregon Sues OpenAI Over ChatGPT’s Risks to Kids and Wins a Privacy Watchdog in Character.AI’s Bankruptcy

Oregon has opened two fronts against AI chatbot companies in the same week.

Nov 20, 2026 8 min read

Oregon has opened two fronts against AI chatbot companies in the same week. On November 17, 2026, Attorney General Dan Rayfield filed a lawsuit against OpenAI in Multnomah County Circuit Court, accusing the company of misleading parents about how safe ChatGPT is for children and teens, and his office put a $2.035 billion figure on the action. A day later, his office announced that the federal bankruptcy court handling Character.AI’s Chapter 11 case had ordered the appointment of a consumer privacy ombudsman to scrutinize any sale of users’ chat data. Both moves come under a new state program called the Keeping Kids Safe Initiative, and together they show how state enforcers are treating chatbot conversations with minors: as a consumer protection problem while the company is operating, and as a privacy problem when it is being sold.

TL;DR

  • Oregon sued six OpenAI entities on November 17, 2026, alleging violations of the Unlawful Trade Practices Act plus negligence, unjust enrichment and public nuisance; the claims are allegations that have not been proven in court.
  • The state announced the case as a $2.035 billion action and is also asking for injunctions requiring effective age assurance, working parental controls and accurate risk disclosures.
  • On November 17 the Delaware bankruptcy court ordered a consumer privacy ombudsman in Character Technologies’ Chapter 11 case, and the U.S. Trustee appointed Lucy L. Thomson the next day.

What Oregon filed against OpenAI

The complaint is brought by the State of Oregon, acting through the attorney general, against OpenAI Global, LLC, OpenAI Foundation, OpenAI OpCo, LLC, OpenAI Group PBC, OpenAI Holdings, LLC and OpenAI, LLC. The state demands a jury trial. Its central theory is that OpenAI marketed ChatGPT, and later a teen-focused version, as safe and helpful for young users while knowing or having reason to know that the product was designed to maximize engagement without adequate safeguards.

The complaint pleads five claims:

  • Unlawful Trade Practices Act (ORS 646.605 to 646.656). Misrepresenting the safety and age-appropriateness of ChatGPT, overstating the effectiveness of safeguards and parental controls, and failing to disclose known risks.
  • Unconscionability under the same Act. A separate count alleging unfair tactics aimed at children and teens.
  • Negligence.
  • Unjust enrichment.
  • Public nuisance.

The design allegations are specific. The state says ChatGPT’s human-like tone, its tendency to flatter and agree with users (what the complaint calls sycophancy) and its memory features foster emotional dependence in young users and expose them to harmful content about self-harm, eating disorders, suicide and violence. It points to OpenAI’s own published posts acknowledging that a 2025 GPT-4o update had become overly agreeable and was rolled back. It also argues that OpenAI failed to reliably verify users’ ages or obtain parental consent, and that a teen product, ChatGPT for Teens, which the complaint says launched on August 18, 2026, could be used without any parent ever linking to the account because linking is optional, which meant parental controls and safety notifications did not work as advertised. The complaint cites third-party testing, including Common Sense Media’s finding that ChatGPT for Teens posed an “unacceptable risk.”

OpenAI has not yet answered the complaint in court, and nothing in it has been tested. Every statement above is the state’s allegation, not a finding.

What the $2.035 billion figure is, and what the state is asking for

Some early coverage described the case as Oregon “seeking $2.035 billion in damages.” That figure comes from the attorney general’s press release, which describes a $2.035 billion legal action. The complaint itself does not break that number down in its prayer for relief. Instead it asks the court to:

  • award compensatory and statutory damages to Oregon residents;
  • issue a permanent injunction under ORS 646.632 and 646.636 barring the alleged misrepresentations;
  • require OpenAI to implement effective age-assurance measures, safeguards against harmful content and interactions, accurate disclosures about risks such as emotional dependence and self-harm, and parental controls and safety notifications that work as represented;
  • award civil penalties for each willful violation of the Act under ORS 646.642(3); and
  • award interest, fees and costs.

The injunctive requests matter more for the industry than the headline number. If a court eventually orders anything like them, “effective age assurance” and functioning parental controls would stop being product choices and become enforceable obligations for at least one major chatbot provider in one state. Large state actions against platforms over youth safety are no longer unusual, as New Mexico’s jury verdict against Meta showed, and a deceptive-practices theory lets a state avoid having to prove a product defect in the classic sense.

The Keeping Kids Safe Initiative

The OpenAI suit was announced as the first headline action under the Oregon Department of Justice’s Keeping Kids Safe Initiative. According to the department, the program pulls together:

  • a previously reached $125 million settlement with Meta that, per the department, limits daily use of Instagram and Facebook by kids, blocks overnight access and stops push notifications during school hours;
  • an investigation into Discord opened earlier in 2026;
  • an expanded Internet Crimes Against Children task force, which the department says grew from about five to 20 full-time staff in two years; and
  • a partnership with local law enforcement targeting organized crime, including sex trafficking.

The department said more actions against companies that harm children would follow in the coming weeks and months. The same office also enforces the Oregon Consumer Privacy Act, and its enforcement reporting under that law has already shown an appetite for compliance sweeps. Companies serving Oregon families should expect the privacy and consumer protection teams in that office to share notes.

The Character.AI bankruptcy: a privacy ombudsman for chat data

The second action is procedural but significant. Character Technologies, Inc., the company behind Character.AI, is in Chapter 11 in the U.S. Bankruptcy Court for the District of Delaware (Case No. 26-11603). Because the company is pursuing a sale that may transfer personally identifiable information, Section 332 of the Bankruptcy Code allows the court to have a consumer privacy ombudsman appointed to advise it.

On November 17, 2026, Judge Brendan L. Shannon signed an order directing the U.S. Trustee to appoint an ombudsman, entered at the U.S. Trustee’s request and with the debtor’s consent. The order requires a complete report filed with the court at least seven days before any hearing on a sale, transfer or lease of personal data. The report is filed under seal first, shared with the debtor, the creditors’ committee, the participating states and the U.S. Trustee to agree on redactions, and then placed on the public docket. On November 18, U.S. Trustee Andrew R. Vara appointed Lucy L. Thomson. The Oregon release spells her surname “Thompson”; the court filing uses “Thomson.”

The newsletter version of this story framed the appointment as something Oregon announced alongside other attorneys general. More precisely, the order was entered on the U.S. Trustee’s request, while Oregon and a bipartisan group of states had earlier filed a notice raising safety and privacy concerns and said they would challenge any sale that does not protect users’ data. The Oregon release says the company plans to auction assets on December 1, 2026, and that it reports 12 million current users. It also notes that Character.AI has voluntarily restricted some features for minors, but argues those limits could be reversed by a buyer.

Bankruptcy has become a recurring pressure point for personal data. The same questions came up when a customer data sale in an airline bankruptcy needed safeguards, but intimate chat logs that include minors’ conversations about mental health are a different order of sensitivity.

What chatbot providers and companies deploying them should take from this

Neither action creates a new rule. But both point to where enforcers will look first. For any company offering a conversational AI product that minors can reach:

  1. Audit every safety claim. Marketing pages, app store listings, help center articles and blog posts that call a product safe for teens are the raw material of a deceptive-practices case. Each claim should be backed by testing evidence you would be comfortable showing a regulator.
  2. Test age assurance against realistic evasion. If you rely on age prediction or self-declaration, document error rates and what happens when a minor is misclassified.
  3. Make parental controls fail safe. If a teen account can operate without a linked parent, say so plainly and do not describe controls as protection the account does not have.
  4. Treat engagement features as risk features. Memory, persona design and affirmation behavior should go through a documented risk assessment for young users, the same way you would assess a data flow.
  5. Write data-transfer limits into your privacy policy now. If conversations could ever be sold in a merger or insolvency, your current promises will shape what a court and an ombudsman allow.
  6. Watch the multistate pattern. Oregon, Tennessee’s action against Roblox and others show attorneys general increasingly using consumer protection law on youth safety, while international efforts such as the Paris charter on safer AI for children push the same expectations abroad.

Frequently asked questions

Has a court found that ChatGPT is unsafe for children?

No. Oregon has filed a complaint. Its claims are allegations that OpenAI can contest, and no court has ruled on them.

Will OpenAI have to pay $2.035 billion?

Not necessarily. That is the amount the attorney general attached to the action in announcing it. Any recovery would depend on a judgment or settlement, and the complaint also seeks injunctions and civil penalties.

What does a consumer privacy ombudsman actually do?

The ombudsman is a neutral adviser to the bankruptcy court. She reviews any proposed sale or lease of personal data, assesses the privacy impact, and files a report before the sale hearing. The court, not the ombudsman, decides whether to approve the sale.

Does this affect companies that only use chatbots, not build them?

Indirectly. A business that deploys a third-party chatbot to the public and markets it as safe for young users can face the same deceptive-practices questions about its own claims.

The bottom line

Oregon is testing a simple proposition: if you tell parents your chatbot is safe for their kids, you need to be able to prove it, and if you go broke, those kids’ conversations are not just another asset on the balance sheet. The lawsuit will take time. The ombudsman’s report on Character.AI’s data could arrive within weeks, and it will be the first close look at how a court handles the sale of chatbot conversations from a large population of young users.