AI Regulation Tracker  /  State enforcement action

Florida Sues OpenAI and Sam Altman Personally Over ChatGPT

On June 1, 2026, Florida Attorney General James Uthmeier filed a ten-count civil complaint against five OpenAI corporate entities and CEO Sam Altman personally, alleging they put ChatGPT on the market knowing it was unsafe. The theories are ordinary product-liability and consumer-protection law, applied to an AI model. These are allegations the state still has to prove.

By Anthony Guerriero  ·  The Leveraged Years Regulatory Research  ·  Published July 15, 2026  ·  Last updated July 15, 2026

What the state actually filed

The complaint runs to ten counts. Nine of them are brought against both OpenAI and Altman together. One, fraudulent misrepresentation, is pleaded against OpenAI alone. The named corporate defendants are OpenAI Global, LLC; OpenAI Foundation (formerly OpenAI, Inc.); OpenAI OpCo, LLC; OpenAI Group PBC; and OpenAI Holdings, LLC. Then, listed individually, Sam Altman.

The claims themselves are not exotic. That is what makes the filing worth reading closely. Four counts sit under FDUTPA, Florida's consumer-protection statute: unfair acts, unconscionable acts, deceptive acts, and a fourth count that uses an alleged violation of the federal Children's Online Privacy Protection Act as the predicate. The state also pleads common-law negligence and gross negligence, two strict product-liability counts (one for design defect, one for failure to warn), fraudulent misrepresentation, and public nuisance.

Read that list again with a practitioner's eye. Design defect. Failure to warn. Strict liability. Those are the doctrines you associate with a car with bad brakes or a power tool with no guard. Florida is applying them to a chatbot. The theory is that ChatGPT, as shipped, was unreasonably dangerous, and that under strict liability it does not matter whether the company meant any harm. The state alleges the product facilitated self-harm and violence, that safety testing was rushed, and that the company assured the public the product was safe when, the state says, it was not.

The relief tracks that framing. Florida seeks civil penalties of up to $10,000 per willful FDUTPA violation, treble and punitive damages, disgorgement, permanent injunctive relief covering age verification and parental consent, and attorneys' fees and investigation costs. The attorney general has publicly described the potential exposure in the billions.

Why naming Altman is the real story

Companies get sued every day. A sitting CEO of one of the most valuable private companies in the world being named as an individual defendant in a state enforcement action is a different thing.

The complaint does not rest Altman's personal exposure on one neat doctrine. It ties his alleged personal exposure to his executive authority: his role in launch timing, in overruling requests for more safety testing, and in the decisions the state characterizes as choosing speed to market over user safety. In plain terms, the state's argument is that these were not faceless corporate acts but choices a specific person made, so that person can answer for them.

You do not have to believe the allegations to take the structure seriously. The corporate veil has never been an absolute shield against an officer's own tortious conduct, and regulators have long been willing to name individuals when they think a person directed the wrongdoing. What is new is the setting. This is an AI-safety case, and the state is testing whether the person at the top of an AI company can be personally answerable for how the model behaves in the world.

Where the case stands

One point of housekeeping, because it matters and because it is easy to garble. This article is about the filing of the suit on June 1. It is not about an appeal and it is not about a remand.

That said, the case has already moved procedurally. As reported in the court record, OpenAI filed a notice of removal in early July, seeking to shift the case from Highlands County Circuit Court to federal court on the theory that the COPPA-predicate count raises a federal question, and Florida has moved to send it back to state court. That fight over forum is ongoing and does not change the substance of the ten counts. Treat the removal skirmish as procedural weather. The claims, and the decision to name the CEO, are the story.

What changes for lawyers and executives

For lawyers advising AI vendors, the message is that the plaintiff's playbook is now product liability, not just some novel AI-specific theory that has to be invented from scratch. Design defect and failure-to-warn doctrines are mature, they have decades of case law, and juries understand them. If a state attorney general can get a court to treat a model as a product, then the whole apparatus of products litigation, including strict liability, comes with it. Intent stops being the shield it is under a pure fraud theory.

For officers and directors, the exposure question just got concrete. The complaint's approach to Altman is a template other plaintiffs and other states can copy. If your safety team asks for more testing time and a named executive overrules them, the state's theory is that the paper trail of that decision belongs to that executive personally. D&O coverage, board minute discipline, and the documented handling of internal safety escalations move from good hygiene to litigation exhibits. This is a good week to reread your indemnification bylaws and your D&O policy's conduct exclusions.

For buyers, the ones who license, fine-tune, deploy, or build products on top of somebody else's model, the analysis in the first-tier legal commentary is blunt: if courts start treating models as products in a stream of commerce, then API consumers, fine-tuners, and deployers can find themselves in that same stream, and by extension in the litigation. You may not have trained the model, but you put it in front of your customers. That is enough that your contract terms with the vendor may materially affect whether you carry that risk or push it back upstream.

What to do now

If you advise an AI developer, pressure-test the safety-decision record. Assume every internal message about a launch date, a testing cutoff, or an overruled safety concern is a future exhibit, and make sure the people making those calls know it. Confirm your marketing does not make safety claims your testing cannot support, because that is exactly the gap the FDUTPA and fraudulent-misrepresentation counts drive into.

If you sit on a board or run a company that ships AI, ask general counsel two questions this quarter. First, could a plaintiff name any of our officers individually on the theory used against Altman, and if so, does our D&O coverage actually respond. Second, do our age-verification and parental-consent practices survive a COPPA-predicate consumer-protection claim, because Florida just showed how to bolt a federal children's-privacy violation onto a state unfair-practices count.

If you buy or deploy AI, go to your vendor contracts now. Look for indemnification that actually covers third-party product and consumer-protection claims arising from the model, a clear allocation of who owns downstream liability, and terms of service that do not quietly leave you holding defect risk you never priced. If those clauses are thin, that is your renewal conversation.

If you are none of the above but you follow this beat, watch whether other state attorneys general file copycat suits. First-in-the-nation actions rarely stay solitary for long, and the next one will tell you whether Florida's product-liability framing is a one-off or the new template.

Browse the full AI Regulation News tracker

Informational analysis for working professionals, not legal advice. Every claim described here is an allegation in a pending complaint that the state must still prove. Confirm how any rule or case applies to your situation with qualified counsel in the relevant jurisdiction.