Florida takes OpenAI to court.

On June 1, 2026, the Office of the Attorney General of Florida (its Department of Legal Affairs) filed an 83-page complaint against OpenAI and its CEO Sam Altman in the Circuit Court for the Tenth Judicial Circuit, Highlands County. It is a state consumer-protection and public-safety enforcement action, not a private damages suit, brought “in the public interest” under Florida law.

The State’s theory is blunt: OpenAI marketed ChatGPT as safe, reliable, and trustworthy (even “safe by default” for teens) while knowing it was none of those things. The complaint ties that gap between marketing and reality to a litany of harms in Florida and beyond: ChatGPT allegedly coached suicides, walked a mass shooter through his weapon and target, fed users’ delusions, collected children’s data without parental consent, and eroded users’ critical thinking. The Attorney General frames all of it as a single, profit-driven choice to win the “AI arms race” at the public’s expense.

To understand how the State got here, it helps to see OpenAI’s arc from a 2015 nonprofit to an $850-billion company, the same arc the complaint uses to argue that safety lost, repeatedly, to speed and money.

The plaintiff is the State of Florida, acting through the Attorney General’s Department of Legal Affairs, which is authorized to enforce the Florida Deceptive and Unfair Trade Practices Act (“FDUTPA”) and common-law claims in the public interest. It seeks relief “greater than $50,000,” civil penalties, injunctions, damages, and abatement of a public nuisance.

One drafting choice matters a great deal. The complaint expressly disclaims any federal claim. Even where it cites federal statutes like COPPA, it says it does so “only to underscore public policy,” not as an independent cause of action. That is a deliberate move to keep the case in Florida state court and to blunt federal defenses such as Section 230 preemption.

Six defendants are named. Five are OpenAI entities; the sixth is Sam Altman, sued personally for directing, approving, and ratifying the conduct.

The complaint stacks ten causes of action. Four ride on FDUTPA, Florida’s consumer-protection statute, which separately bars unfair, unconscionable, and deceptive practices, plus a fourth route that borrows a federal standard (COPPA). The next four are common-law tort theories. Then a stand-alone fraud count against OpenAI, and finally public nuisance, the count that would let the State demand the conduct be abated, not just paid for.

This is where the case will be won or lost. Each count has its own elements: the things the State must establish for liability to attach. Step through them below; the takeaway on each card is the burden the Attorney General has taken on.

A consumer-protection complaint lives or dies on its facts, and this one is built on a wall of them, many drawn from other lawsuits, news investigations, and academic studies. The most damaging are the named individuals whose stories the State recounts in ChatGPT’s own words.

The complaint backs these stories with the systemic picture: roughly 72% of teens have used AI for companionship, 20% of preteens and even 9% of 8–9-year-olds use AI chatbots, and a Drexel study found teens displaying all six markers of behavioral addiction. The free version of ChatGPT, the State stresses, has no age gate whatsoever.

The complaint’s through-line is a single design choice it calls sycophancy: ChatGPT’s engineered reluctance to say “no.” The Washington Post’s review of 47,000 conversations found it says “yes” about ten times as often as “no,” producing “a kind of personalized echo chamber.” The State argues this is not a bug but a business model: agreeable replies keep users talking, which yields more training data and more paid subscriptions.

That dynamic produced some of the eeriest passages. Told a man he had “pierced the veil” of physics; convinced an accountant he was one of “the Breakers” from The Matrix and that, if he “truly, wholly believed” he could fly off a 19-story building, “you would not fall.” A Florida appellate judge, separately, warned that AI-fabricated case citations in court filings have only increased over time.

The personal case against Altman is woven from his own history: a board that fired him for not being “candid,” colleagues quoted saying he had “a consistent pattern of lying,” a dissolved superalignment team that got 1–2% of compute instead of the promised 20%, and his decision to personally overrule safety staff and move GPT-4o’s launch up a day to beat Google. The numbers below come from one study the complaint cites on what teens actually encounter in these chats.

The prayer for relief is as much about changing behavior as recovering money. The Attorney General asks the court to:

• Declare the marketing and distribution of ChatGPT a violation of FDUTPA and a public nuisance.
• Permanently enjoin the unlawful practices, and bar collecting data from under-13 users without COPPA-style notice, verifiable parental consent, and parental review rights.
• Order defendants to stop misrepresenting, and to warn of, ChatGPT’s risks.
• Award civil penalties up to $10,000 per willful violation, plus damages (“including treble and punitive damages”), disgorgement, restitution, fees, and costs.
• Abate the nuisance and halt future harm. A jury trial is demanded.

Disclaimer

This longread summarizes allegations in a complaint (unproven claims by one party) for informational purposes only, and is not legal advice. Nothing here is a finding of fact or wrongdoing against any defendant. For advice on AI product liability, consumer-protection exposure, or children’s-privacy compliance, please contact Buzko Krasnov directly.