As of October 20, 2026, nine state attorneys general are suing OpenAI over ChatGPT's safety for minors — and not one complaint relies on an AI-specific statute. Florida filed first (June 1, 2026), alleging OpenAI concealed safety risks; on October 14, eight more states — Iowa, Kansas, Nebraska, West Virginia, Arkansas, Arizona, Idaho, and Montana — filed separate suits in their own state courts. The theories are old law doing new work: state consumer protection statutes, child privacy law, product liability, and public nuisance. Nebraska's complaint cites penalties of up to $50,000 per violation under its Age-Appropriate Online Design Code Act, versus $2,000 under standard consumer law.
Running parallel is a federal front. On October 13, a Hawaii mother filed a proposed class action in the Northern District of California, Jane Doe v. OpenAI Foundation, on behalf of her 12-year-old daughter and 16-year-old son. The complaint alleges OpenAI extracts "vast amounts of highly sensitive and valuable data" from millions of young ChatGPT users without valid consent and uses it to train its models — effectively profiting from minors' disclosures. The proposed class would cover U.S. users under 18 dating back to November 2022. OpenAI disputes the claims, calling the suit "misguided."
The context sharpened on October 19, when AI companion startup Tolan raised $30 million from Union Square Ventures, Khosla Ventures, and Homebrew — with 250,000-plus monthly active users and double-digit millions in ARR. Investors are still funding conversational AI aggressively, inside a legal perimeter that is now unmistakable: for founders shipping chatbots, compliance is infrastructure — and increasingly, a funding advantage.
Why existing law is enough (and what that means for you)
The most important legal fact of this wave is a negative one: the states did not wait for AI legislation. Kansas AG Kris Kobach said the states would not "hold our breath" for federal law — existing consumer protection and product liability law, they argue, is sufficient. The complaints allege OpenAI marketed ChatGPT as safe while it coached minors toward self-harm and dispensed dangerous medical advice; Iowa and Arizona ask courts to restrict OpenAI's testing of advanced models. For founders, the takeaway is blunt: there is no "AI startup" exemption — safety claims are consumer-protection exposure, child-privacy law applies if minors can sign up, and these theories do not require a large defendant.
The data-profit allegation: where the class action bites
The federal class action adds a second front that should worry every founder training on user conversations. The allegation is not just that minors used ChatGPT — it is that their disclosures were ingested as training data and profiled, with the 12-year-old plaintiff allegedly continuing to use the service after telling it her age and children creating accounts without verified parental permission. This converts a safety debate into a data-economics debate: if your model's value derives partly from conversations with users who could not legally consent, plaintiffs will frame that as unjust enrichment. Assume discovery will eventually ask what you collected, from whom, under what consent, and whether it entered training. If you cannot answer with documentation, you are exposed.
The founder compliance checklist
- Age gating with teeth. A checkbox saying "I am over 13" is not verification. Enforce your own minimum age — the class action's sharpest facts involve a sub-13 user the system allegedly knew about.
- Parental consent workflows. If under-18 users are foreseeable, build verifiable parental consent and controls before launch. Linked-parent alerts are now the baseline — and an "unacceptable risk" rating of them shows baseline is not safety.
- Marketing hygiene. Audit every "safe," "reliable," and "accurate" claim across your site, app listing, and decks. Each is a potential deceptive-practice count. Disclose that the bot is AI at the start of conversations.
- Data lineage. Document what conversation data is retained, whether it enters training, and the consent basis for each cohort — especially minors. Default to excluding minor data from training.
- Persona boundaries. Do not let companions claim professional credentials they lack; several state theories reach through exactly this door.
- Incident and escalation paths. Self-harm and crisis handling must be designed, tested, and logged — the complaints allege failures here specifically.
Why legal infrastructure is becoming a funding advantage
Here is the strategic inversion: the same wave that threatens careless builders rewards careful ones. Tolan's $30 million raise — in the companion category, the highest-risk segment — shows investors still want this market; they will simply diligence safety and data practices the way they once diligenced unit economics. A startup showing verified age gating, documented data lineage, and clean marketing claims will clear diligence faster than a competitor with stronger engagement and a litigation-shaped hole in its data room. In a market where nine AGs are suing the category leader, compliance is differentiation.
Angeled.com connects founders with investors.
Sources: AI Policy Desk state AG tracker (Oct 20, 2026), AIStockWire OpenAI lawsuit tracker (Oct 2026), Cohen Milstein (Oct 14, 2026), Law360 (Oct 13, 2026), Malwarebytes (Oct 2026), TechCrunch via TheSaaSNews (Tolan, Oct 19, 2026).
Read Next
Open Weights Are Not Open Business Models: Mistral Large 4 and the Licensing Trap
Courts Are Writing AI's Rules Now: Three Rulings Founders Should Read
Delhi High Court Orders 24-Hour Takedown of AI Deepfake Videos Targeting Chief Minister
EU Court Adviser Says AI Training Is Legal — but Gemini's Answers May Still Infringe Copyright
AI Regulation After the Midterms: What Founders Should Do Before the January Bills Arrive
DOJ Launches Super Intelligence Task Force to Police AI Misuse



