By Jeffrey T. Donner, Esq.

June 2, 2026

Florida has sued OpenAI and Sam Altman over ChatGPT.

The lawsuit is being sold as consumer protection, child protection, and public safety. But the deeper problem is not artificial intelligence. The deeper problem is what this lawsuit says about our legal culture.

We are drifting very far from the legal system we were all taught this country was built upon.

We were all taught that the American system protects people through limited government, private property rights, freedom of contract, open markets, divided power, and courts that decide actual legal disputes. We were taught that legislatures make policy, courts apply law, and adults are generally free to choose which lawful tools, products, services, and technologies they want to use.

Some people apparently did not absorb the lesson. But that is the system.

The Founding generation understood that protection of private property, contract, enterprise, and economic liberty was not a side issue. It was central to constitutional government. Alexander Hamilton, John Jay, John Marshall, and the lawyers and statesmen of that era understood that a government of separated powers and enforceable rights exists in large part to protect citizens from arbitrary government interference with their property, contracts, businesses, and lawful private choices.

That does not mean fraud is allowed. It does not mean companies can deceive consumers. It does not mean children cannot be protected. It does not mean dangerous conduct cannot be regulated.

But the baseline is supposed to be liberty.

This lawsuit reflects a different instinct: the government identifies a modern anxiety, labels it a public danger, invokes children, stretches consumer-protection law and public-nuisance doctrine, and then asks a court to regulate a general-purpose technology used by millions of people.

That is not limited government. That is not the free market. That is not the common law in any disciplined sense.

It is government-by-lawsuit.

I. Courts are not technology regulators.

The most basic problem is institutional.

Courts decide cases. They apply statutes. They resolve concrete disputes. They determine whether a defendant violated a recognized legal duty and, if so, what remedy the law permits.

Courts are not supposed to become statewide technology regulatory agencies.

If artificial intelligence needs regulation — and some aspects of it probably do — the Legislature can address that. Legislatures can hold hearings. They can define terms. They can hear from parents, engineers, businesses, teachers, lawyers, doctors, privacy experts, child-safety advocates, and ordinary users. They can decide what conduct should be prohibited, what disclosures should be required, what remedies should exist, and how to balance risk against lawful use.

That is legislation.

A lawsuit is not legislation. A civil complaint is not a regulatory code. A circuit judge is not an AI commissioner.

The State’s lawsuit asks a court to do something courts are poorly suited and constitutionally ill-suited to do: supervise or reshape a general-purpose artificial-intelligence tool because the government says the tool creates broad social risks.

A court can enjoin a proven statutory violation. A court can stop a specific deceptive practice. A court can award damages based on evidence. A court can apply law to facts.

But a court should not be asked to manage the relationship between Floridians and artificial intelligence.

That is not judging. That is governing.

II. This is regulation through litigation.

One of the reasons lawsuits like this are attractive to government is that they avoid the hard work of legislation.

Legislation requires debate. It requires votes. It requires definitions. It requires compromise. It requires political accountability. Legislators have to say what the rule is, who it applies to, what conduct is forbidden, what exceptions exist, and what remedies are available.

A lawsuit can avoid all of that.

The State can file a dramatic complaint. It can assemble worst-case examples. It can invoke children, safety, addiction, mental health, deception, and public harm. It can use broad words like unfair, deceptive, dangerous, addictive, and nuisance. Then it can ask a judge to impose sweeping relief.

That is much easier than passing a carefully written AI statute.

It is also much more dangerous.

If the government wants to regulate AI, it should do so openly, legislatively, and prospectively. It should not use a lawsuit to ask one judge to become the regulator of a technology used by lawyers, businesses, students, teachers, parents, workers, and ordinary citizens.

That is not how separation of powers is supposed to work.

III. Public nuisance does not mean “the government does not like this.”

The public-nuisance theory is especially troubling.

Nuisance is an old common-law doctrine. It has a real place in the law. Lawyers learn nuisance through concrete examples: a smokestack sending soot onto neighboring property; a factory polluting a river; dust, fumes, noise, odors, flooding, blocked access, or some other interference with public rights or the use and enjoyment of property.

That is nuisance.

Nuisance does not mean: “The government dislikes this product.”

It does not mean: “This technology makes public officials anxious.”

It does not mean: “Some people may misuse this tool.”

It does not mean: “The State believes this lawful product creates broad social risks.”

And it certainly should not mean: “A court may regulate an entire industry because the plaintiff has used the word nuisance.”

That is not common-law development. That is common-law abuse.

Oliver Wendell Holmes famously wrote that the life of the law has not been logic, but experience. That is a profound statement about how the common law develops. The common law grows from real disputes, practical experience, analogy, custom, reason, and disciplined judicial decision-making. It does not grow by taking an old doctrine, tearing it away from its historical setting, and using it as a modern policy weapon whenever a plaintiff wants a court to solve a social problem.

Common-law creativity has limits.

A judge deciding a nuisance case is not supposed to become a legislature. A nuisance claim is not supposed to become a regulatory scheme. And the common law is not supposed to be an all-purpose invitation for courts to manage whatever new product or technology the government has decided to target.

If ChatGPT is a public nuisance because it can be misused, then almost anything can be a public nuisance.

Search engines can be nuisances because people search for harmful information. Smartphones can be nuisances because people become addicted to them. Social media can be a nuisance because minors use it too much. Video games can be nuisances because some people blame them for violence. Books can be nuisances because unstable people can claim they were influenced by them. Cars can be nuisances because crashes happen. Fast food can be a nuisance because obesity exists.

And alcohol is the most obvious example.

Alcohol is a lawful product. It is also involved in enormous social harm. People abuse it. People become addicted to it. People drive drunk. People die. Innocent third parties die. Families are destroyed. Children are harmed. Crimes are committed. Hospitals, police departments, courts, and taxpayers all bear costs connected to alcohol abuse.

Under the logic of Florida’s lawsuit against OpenAI, why should the Attorney General not sue every distillery, brewery, liquor store, bar, restaurant, grocery store, and convenience store that sells alcohol? If a lawful product can be treated as a public nuisance because some users misuse it and serious downstream harm occurs, then alcohol is a far stronger candidate than ChatGPT.

But that is not how law is supposed to work.

We regulate alcohol through statutes, licensing, criminal law, age restrictions, DUI laws, dram-shop principles where applicable, and specific civil claims based on specific duties. We do not convert the entire alcohol industry into a public nuisance simply because alcohol can be misused and can cause death.

That distinction matters.

A lawful product does not become a nuisance merely because some users abuse it, misuse it, misunderstand it, overuse it, or use it in connection with harmful conduct. If that were the rule, almost every useful but risky product in America would be subject to nuisance litigation.

Once nuisance means “this lawful product contributes to a broad social problem,” there is no limiting principle.

That is not nuisance law. That is government-by-litigation with a common-law label pasted on top.

IV. FDUTPA is not a blank check to regulate new technology.

Florida’s Deceptive and Unfair Trade Practices Act gives the Attorney General broad authority. That is not seriously debatable.

But broad authority is not unlimited authority.

FDUTPA is supposed to address unfair or deceptive acts or practices in trade or commerce. If OpenAI made a specific safety representation that was false when made, plead that. If OpenAI violated a specific privacy law involving minors, plead that. If OpenAI deceptively marketed a particular feature, identify the feature, the statement, the deception, the consumers affected, and the injury.

That is consumer-protection law.

What the State should not be allowed to do is use FDUTPA as a back door to put an entire technology on trial.

A consumer-protection statute is not an AI regulatory code. It is not a general license for the State to ask a court to supervise the design, warnings, outputs, training, availability, and user experience of a general-purpose information tool.

Remedies follow violations. They do not create policy power.

If the State proves a specific deceptive practice, the court can address that practice. But the court should not let FDUTPA become a vehicle for judicial management of artificial intelligence.

V. The lawsuit is part of a larger drift away from freedom of contract and private property.

There is a larger problem here, and it is not limited to AI.

We are living in a legal culture increasingly distant from the world of private property, freedom of contract, and limited government that this country was founded to protect.

Private parties enter contracts, build tools, sell services, assume risks, compete in markets, make choices, and bear consequences. That does not mean fraud is allowed. It does not mean companies can deceive consumers. It does not mean children cannot be protected. It does not mean dangerous conduct cannot be regulated.

But the baseline is supposed to be freedom.

Adults are supposed to be able to choose lawful products. Businesses are supposed to be able to innovate. Consumers are supposed to be able to decide what works for them. Competitors are supposed to discipline bad products. Markets are supposed to reward better alternatives. Courts are supposed to enforce rights, not manage life.

That structure has eroded.

Increasingly, when government officials dislike something, they do not merely regulate specific unlawful conduct. They try to control the product itself. They try to turn social discomfort into legal authority. They try to convert anxiety into standing, policy into nuisance, and political theater into judicial relief.

That is not the free-market system this country was built on.

It is not freedom of contract.

It is not respect for private property.

It is not respect for private enterprise.

It is not respect for private decision-making.

It is the government saying: we know better; we will decide; the court will enforce our view.

That is exactly the wrong instinct in a free society.

VI. The COVID lesson should have been that emergencies do not erase legal limits.

The pattern is familiar because we saw it during COVID.

Government identifies a serious risk. Government invokes safety. Government insists that the ordinary rules must bend. Government demands deference. Businesses close. Churches are restricted. Schools shut down. Landlords are told they cannot enforce leases. Employers are told how to manage their workplaces. Citizens are told that constitutional limits and private choice must yield because the government has declared an emergency.

Some COVID risks were real. Some government responses were defensible. But the broader lesson should have been clear: serious problems do not erase legal limits.

The Supreme Court eventually pushed back in cases like Roman Catholic Diocese of Brooklyn v. Cuomo, Alabama Association of Realtors v. HHS, and NFIB v. OSHA. Those cases were not about AI, but they stand for a principle that matters here: the government does not gain unlimited power merely because it identifies a serious public problem.

Florida’s lawsuit against OpenAI reflects the same bad habit.

The State identifies real anxieties about AI: children, mental health, misinformation, addiction, deception, data, safety. Then it tries to convert those anxieties into power for a court to regulate a general-purpose information tool.

That is the same pattern.

Different subject. Same instinct.

VII. “Protect the children” cannot become a password for unlimited state power.

No serious person is against protecting children.

Children should be protected from exploitation, fraud, abuse, unlawful data collection, and manipulative design. Parents should have meaningful tools. Companies should not lie about safety. If OpenAI violated specific legal duties involving minors, those violations can be pleaded and litigated.

But “protect the children” cannot become a password for unlimited government power.

That phrase has been used for decades to justify restrictions on new media and new technology: comic books, rock music, movies, video games, the internet, social media, and now artificial intelligence.

Sometimes the concern is partly legitimate. Sometimes it is exaggerated. Often it is both.

But the legal answer cannot be: because children might be harmed, the government may regulate the tool for everyone.

Parents have a role. Markets have a role. Schools have a role. Professional standards have a role. Legislatures have a role. Courts have a role when actual legal violations are pleaded and proven.

But child-protection rhetoric does not dissolve separation of powers.

It does not eliminate statutory limits.

It does not eliminate causation.

It does not eliminate constitutional concerns.

And it does not turn a trial judge into a technology regulator.

VIII. This is not limited-government litigation merely because the target is Silicon Valley.

There is an obvious irony here.

For years, conservatives criticized progressive lawyers and Democratic officials for using courts to obtain policy outcomes they could not get through legislatures. They criticized regulation through litigation. They criticized public-nuisance theories used to attack entire industries. They criticized efforts to turn judges into policy managers.

Often, those criticisms were correct.

But now the same method is being used under different branding.

The target is not an oil company, gun manufacturer, pharmaceutical company, or landlord. The target is a Silicon Valley AI company. The slogans are not climate, equity, corporate greed, or public health. The slogans are children, AI safety, deception, addiction, and public harm.

But the legal move is the same.

Take a broad social concern. Call it nuisance. Add consumer protection. Invoke safety. Ask a court to impose broad control.

That is not limited government merely because Republicans are doing it.

Since COVID, it has become harder to tell the difference between the parties when the question is government power. Both sides now seem comfortable with overreach when the target fits their preferred narrative. The left does it in the name of public health, equity, climate, and corporate accountability. The right does it in the name of children, morality, safety, and cultural protection.

The labels change.

The power grab remains.

If the principle is limited government, then it has to apply even when the government is attacking a company that is politically easy to attack.

IX. The First Amendment and information-tool problem cannot be ignored.

ChatGPT is not merely a physical product.

It is an information tool. It generates words. It responds to questions. It drafts, summarizes, organizes, translates, explains, compares, analyzes, and produces ideas. It is software, but it is also a speech-generating and information-processing system.

That does not make OpenAI immune from law. It does not mean companies can lie. It does not mean children’s data can be mishandled. It does not mean every AI output is protected in every possible context.

But it does mean courts should be careful before allowing the State to regulate an information-generating technology through broad consumer-protection and nuisance theories.

The Supreme Court has repeatedly been skeptical of government efforts to restrict new forms of expression or communication because officials believe those forms may harm children. In Brown v. Entertainment Merchants Association, the Court struck down California’s attempt to restrict violent video games to minors. The Court did not hold that video games are harmless. It held that government cannot bypass constitutional limits by declaring a new medium dangerous.

Artificial intelligence is different from video games, but the caution is similar.

New information technologies always create fear. The printing press, radio, television, the internet, search engines, smartphones, social media, and now AI all created new risks.

A free society does not respond by letting the State decide which information tools adults may use.

X. Lawful users are not collateral damage.

One of the major defects in broad technology lawsuits is that they ignore lawful users.

That matters here.

I use ChatGPT in my law practice. Not as a substitute for legal judgment. Not as a replacement for research. Not as an authority. Not as a lawyer. Not as a court. Not as a source to trust blindly.

I use it as a tool.

It helps organize ideas. It helps create first drafts. It helps test arguments. It helps revise language. It helps identify issues. It helps turn rough thoughts into structure. Every output still has to be checked, edited, verified, corrected, and owned by the human lawyer.

That is responsible professional use.

Millions of people use AI responsibly too. Students use it to learn. Businesses use it to plan. Workers use it to write and organize. Parents use it to explain things. Disabled users use it for communication and access. Non-native English speakers use it to write more clearly. Ordinary citizens use it to understand documents and prepare better questions.

Those people do not disappear because the State has collected worst-case allegations.

A statewide injunction against a general-purpose technology affects lawful users. Courts should be skeptical of remedies that treat lawful adult use as collateral damage in a government lawsuit.

XI. The free market is not the enemy. It is part of the protection.

The free market is not perfect. No serious person thinks it is.

But the free market is part of the American system of protection.

If OpenAI makes a bad product, users can leave. Competitors can build better tools. Businesses can adopt stricter policies. Parents can restrict use. Schools can regulate students. Lawyers can develop professional standards. Consumers can demand better safeguards. Legislatures can pass specific laws. Courts can adjudicate actual legal claims.

That is how a free society is supposed to handle technological change.

The alternative is centralized control by lawsuit. The government decides a technology is dangerous, invokes children and public safety, calls the product a nuisance, and asks a court to impose sweeping relief.

That is not market discipline.

That is not limited government.

That is litigation as regulation.

XII. The right legal answer is precision, not panic.

The legal system should not ignore real AI problems.

If there are false safety claims, plead them. If there are privacy violations, enforce the privacy laws. If minors need better protection, pass specific rules. If professional users need standards, bar regulators, courts, schools, and industries can develop them. If a particular design feature is deceptive or dangerous in a legally cognizable way, identify it and prove it.

But do not put the entire technology on trial.

Do not stretch public nuisance beyond recognition.

Do not use FDUTPA as a back door to general technology regulation.

Do not ask a circuit judge to do the Legislature’s job.

Do not let the emotional force of tragic allegations obscure the structural question: does the court have the institutional authority to grant the kind of relief the State is really seeking?

That question should be central.

XIII. Conclusion

Florida’s lawsuit against OpenAI may contain some legitimate claims inside it. If OpenAI made specific false statements, violated specific laws, or deceived consumers in a legally actionable way, those issues can be litigated.

But as a broader matter, the lawsuit appears to ask the judiciary to do something courts are not supposed to do: regulate a general-purpose technology through expansive consumer-protection and public-nuisance theories.

That is dangerous.

Public nuisance should not mean “the government does not like this.”

FDUTPA should not become a technology-management statute.

A circuit court should not become an AI regulatory commission.

And child-protection rhetoric should not be allowed to dissolve separation of powers.

We are very far from the disciplined common law Holmes described. We are also far from the basic civic understanding we were all taught: that America protects people through liberty, private property, freedom of contract, limited government, divided power, competition, and markets — not through government officials asking judges to manage every risk created by every new technology.

Courts decide cases.

Legislatures regulate industries.

Markets discipline products.

Parents make decisions for children.

Adults make decisions for themselves.

That is the system.

If artificial intelligence needs regulation, regulate it lawfully, specifically, and prospectively. Pass statutes. Define duties. Create standards. Enforce actual violations.

But do not ask a trial court to become the branch of government that manages every risk created by a new technology.

That is not judging.

That is governing from the bench.