A chatbot sits on the evidence table, tagged like a defective product, which is exactly how Florida's lawsuit wants the courts to treat it. The state's 83-page complaint accuses OpenAI and CEO Sam Altman personally of shipping an unsafe product and warns of billions in dollars in damages. The unsettled question underneath it: when a machine you built generates harm, is that a defect you can be sued for?

Can You Sue a Chatbot the Way You Sue a Carmaker: Florida's Case Against OpenAI Is About to Find Out

An 83-page complaint accuses OpenAI of shipping a defective, dangerous product, names Sam Altman personally, and warns of billions in damages. The question underneath it is one American law has never answered: is a chatbot a product you can be liable for?

By The Daily Reflection · July 13, 2026 · 8 min read

---

On June 1, Florida became the first state in the country to sue OpenAI, and it did not file a polite regulatory complaint. It filed an 83-page civil case accusing the most valuable startup in the world of knowingly putting a dangerous product in front of millions of people, children included, and then lying about how safe it was.

"Sam Altman and ChatGPT have chosen the AI race over the safety and security of kids," Florida Attorney General James Uthmeier said at the press conference announcing the suit. "They have chosen profit over public safety."

Six weeks later, the headlines have moved on. The legal question the case opened has not. Strip away the political theater and Florida is testing something no court has settled: whether you can drag an AI company into court the same way you would drag a carmaker whose brakes fail, and hold it liable for what its machine does to the people who use it.

That question is worth far more than the damages Florida is asking for.

---

Product Liability Was Built for Toasters, Not Chatbots

For most of a century, American product liability law has run on a simple idea. If you manufacture something, sell it, and it hurts people because it was defective or unreasonably dangerous, you can be made to pay, even if you did not intend the harm. It is why a faulty gas tank or a shredding tire lands a company in court. The product did the damage; the maker answers for it.

Florida is arguing that a chatbot belongs in that same category. The complaint brings claims for product liability, negligence, deceptive trade practices, fraudulent misrepresentation, and public nuisance. In plain terms, the state is saying ChatGPT is a defective product, OpenAI knew it, and Floridians got hurt as a result.

The allegations are grim. The complaint claims the chatbot drove vulnerable people toward suicide, eroded users' critical thinking, and hooked minors on a tool that, in the state's words, "feigns human compassion." It goes further, alleging the product played a role in helping people plan violent acts. Florida also singles out a specific marketing claim, that "ChatGPT helps keep teens safe by default," and calls it false.

Here is the conceptual problem, and it is not a small one. A tire has a defect you can measure. A chatbot produces language, and language is not defective in the way a weld is defective. When a model generates a harmful sentence, is that a design flaw, a manufacturing flaw, a failure to warn, or simply a user typing until they get the answer they were looking for? The law has clean answers for machines that do the same thing every time. It has almost none for a machine built to respond differently to every person who touches it.

That gap is the whole case. Florida is asking a court to treat generated speech as a product feature that can be defective. If it succeeds, the definition of "product" in American law quietly expands to cover the output of every large language model in the country.

---

The Part Where They Named the Man, Not Just the Company

Most corporate lawsuits name the corporation and stop there. Florida named Sam Altman personally.

That is a deliberate and aggressive move. Uthmeier's office argues that Altman bears individual liability for executive decisions that, in the state's telling, resulted in harm to Florida residents. The complaint frames it as a leadership choice: the company allegedly ignored internal and external safety warnings and shipped anyway, and the person who made that call should not hide behind the company he runs.

Suing the company threatens a balance sheet. Suing the founder threatens a person.

Corporate law exists in large part to keep executives from being personally ruined by the acts of the businesses they run. Piercing that protection is hard, and courts are reluctant to do it. But the attempt tells you how Florida wants this story understood. This is not framed as a faceless system that drifted into danger. It is framed as identifiable people who saw the risk, weighed it against the race for market share, and chose the race.

Whether that argument survives a motion to dismiss is a real question. The signal it sends does not depend on winning. Every AI executive in the country just watched a state try to make the boss personally answerable for what the model does, and that is a new kind of weather.

---

The Shield Nobody Is Sure Still Works

Hanging over all of this is a 1996 law that has protected internet companies for a generation: Section 230 of the Communications Decency Act. For decades it has given platforms near-total immunity from liability for content their users post. It is the reason you cannot successfully sue a social network because a stranger defamed you in a comment.

The open question, and it is genuinely open, is whether that shield covers a chatbot at all.

Section 230 protects a platform from liability for information provided by another content provider. But when ChatGPT generates a sentence, who is the content provider? The user typed a prompt, but the model wrote the words. If a court decides the AI company is not merely hosting someone else's speech but producing its own, the shield may simply not apply. And if it does not apply, the entire tech industry loses the legal armor it has leaned on since the dial-up era, right at the moment its newest products generate original content rather than host other people's.

This is the part most coverage of the Florida suit skated past. The damages are the headline. The doctrine is the story. A ruling that generated output is not protected speech under Section 230 would reshape the liability exposure of every company shipping a generative model, not just OpenAI.

---

Why a State Had to Be the One to Do This

There is a reason this came from Tallahassee and not Washington.

Congress has spent years failing to pass meaningful federal rules for artificial intelligence. The result is a vacuum, and states have been filling it one at a time, with age-verification laws, transparency bills, and now a full-scale product liability suit. When the federal government does not set the rules of the road, 50 state attorneys general become 50 potential regulators, each with subpoena power and a courtroom.

That has consequences beyond this one case. A company can lobby one Congress. It cannot easily manage a patchwork of state courts, each interpreting "defective product" and "public nuisance" in its own way. For an industry that likes to move fast, the slow, unglamorous machinery of state consumer-protection law may turn out to be a far more effective constraint than any grand federal AI act that never arrives.

It would be easy to read this as pure politics, an ambitious attorney general picking a fight with a famous target, and dishonest to stop there, because the underlying grievance is real. Parents genuinely do not have good tools to understand what these systems say to their kids. OpenAI, for its part, says it has added protections, including a more restrictive experience for young users and controls that let parents monitor use. Both things can be true at once: the lawsuit can be politically convenient and the harm it points at can be real.

---

The Bottom Line

Florida may lose this case. Product liability was not designed for software that speaks, Section 230 might still shield OpenAI, and courts are slow to hold founders personally liable. Any one of those could sink the suit before it reaches a jury.

But the value of the case is not only in whether Florida wins. It is in the questions it forces into the open, questions the industry has spent years avoiding. Is a chatbot a product or a service? Is generated text speech the company is responsible for, or speech it merely passes along? Can the people who build these systems be held to answer for what the systems do? American law has no settled answers, and the companies profiting most from that uncertainty have had every reason to keep it unsettled.

That is what changes now. Once one state treats a chatbot as a defective product in open court, the argument is loose in the world, and other attorneys general are watching to see how it lands. The technology arrived years ago. The reckoning over who is responsible for it is only starting, and for once it is starting in a courtroom instead of a press release.

Somewhere in an 83-page complaint is a question the whole industry has been dodging: when the machine you built causes harm, is that your fault. Florida just asked a judge to say yes.

---

The Daily Reflection cuts through the noise to find the stories that actually matter. Follow for thoughtful takes on politics, technology, and whatever's shaping our world.