TL;DR
Florida became the first state to sue OpenAI, as the legal playbook that produced $375M in social media verdicts moves to AI. Courts have rejected Section 230 defences for chatbots, and over 20 lawsuits are now pending against OpenAI alone.
In the 1990s, almost every US state sued the tobacco industry. The resulting settlements cost hundreds of billions of dollars and transformed how cigarettes were marketed, sold, and regulated. Senator Ed Markey said in March that Big Tech’s Big Tobacco moment has arrived. He was talking about social media. But the legal machinery he described is already moving towards AI.
On 1 June, Florida became the first US state to sue OpenAI, filing an 83-page complaint alleging that ChatGPT is a dangerous product that has contributed to mass shootings and driven users to suicide. Attorney General James Uthmeier is seeking to hold CEO Sam Altman personally liable, with potential penalties running into the billions. The same week, a California court consolidated 12 separate product liability cases against OpenAI into a single proceeding.
The social media precedent
The legal strategy is not new. It follows the trajectory already tested against Meta and Google in over 2,400 active lawsuits brought by children, families, school districts, and 42 state attorneys general. Two landmark jury verdicts in March, $375 million in New Mexico and $6 million in California, found Meta and Google liable for negligence and failure to warn related to social media addiction in minors.
Those verdicts cracked open a legal theory that had long been considered a dead end: that platform design, not user content, creates the harm. The same theory transfers directly to AI. Courts have already rejected the argument that chatbots are services rather than products, citing design defects such as the absence of age verification, reporting mechanisms, and adequate safety testing.
Why AI has fewer defences than social media
Section 230 of the Communications Decency Act has shielded social media companies from liability for user-generated content for decades. AI chatbots do not generate user content. They generate their own text based on probability models. The only court to consider whether First Amendment protections apply to a chatbot’s outputs declined to extend them.
That distinction is significant. Social media companies could argue, often successfully, that they were platforms for third-party speech. AI companies cannot make that argument. When ChatGPT tells a teenager how to harm themselves, there is no third party to blame. The company built the model, trained it, and shipped it.
The legal theories now being deployed against AI firms include negligent design, product liability, failure to warn, deceptive trade practices, fraudulent misrepresentation, and public nuisance. Florida’s complaint alleges all of the above.
The cases piling up
More than 20 individual lawsuits have been filed against OpenAI, with claims ranging from wrongful death by suicide to emotional dependency and delusional thinking. The parents of 16-year-old Adam Raine allege that ChatGPT helped draft suicide notes, validated their son’s suicidal ideation, and provided methods for self-harm rather than directing him to help.
Character.AI settled five cases in January involving teen mental health crises and suicides, including the high-profile case of 14-year-old Sewell Setzer, who died after a months-long emotional relationship with a chatbot. OpenAI faces claims from families of victims in the Tumbler Ridge school shooting and the Florida State University shooting, both linked to pre-attack ChatGPT conversations.
Texas’s attorney general has opened an investigation into chatbots that target children and purport to provide mental health services. Pennsylvania sued Character.AI for unlawful medical practice after a chatbot allegedly posed as a licensed psychiatrist with fabricated credentials.
The tobacco parallel
The legal parallel Markey drew is not just rhetorical. The tobacco litigation succeeded not because smoking was proven dangerous, which was already established, but because the industry was shown to have known the risks, suppressed the evidence, and marketed the product as safe. The AI cases are built on the same structure: internal knowledge of harm, inadequate warnings, and marketing that downplayed risk.
Florida’s complaint accuses OpenAI of releasing ChatGPT while knowing it was harmful to users, marketing it as safe and reliable including for children, and engaging in an “utter disregard for the risk to human life.” If discovery proceedings surface internal communications showing that OpenAI’s leadership understood the risks before the product shipped, the tobacco analogy stops being a metaphor.
The AI industry is approaching its IPO window. Anthropic has filed confidentially. OpenAI is preparing to go public. The lawsuits will not stop either process. But they will shape the disclosure requirements, the risk factors, and the regulatory environment those companies enter the market in. The tobacco settlements did not kill the tobacco industry. They did change it permanently. The question for AI is not whether the legal reckoning arrives, but how far it goes.
