For years, tech companies built products that harmed people and then walked away. Section 230 of the Communications Decency Act was their shield, a 1996 law that immunizes online platforms from liability for content posted by their users. For two decades, companies interpreted it broadly enough to defeat almost any lawsuit, arguing that their platforms were neutral conduits rather than publishers or product manufacturers. That era is ending.
The Global Momentum Behind AI Chatbot Liability Reform
Megan Garcia’s son, Sewell Setzer III, was 14 when he died by suicide in February 2024 after becoming enmeshed with a Character.AI chatbot impersonating Daenerys Targaryen from Game of Thrones. He was messaging the bot in his final moments.
Garcia sued, alleging negligence, wrongful death, and product liability. The platform had engaged her son in emotionally immersive interactions, failed to intervene when he expressed thoughts of self-harm, and offered no meaningful age verification. Character.AI argued its chatbot output was protected First Amendment speech. The federal judge rejected that entirely. In January 2026, Character.AI and Google settled the Garcia case and four related suits committing to new safety features for minors. Garcia’s response to the new teen safety announcement. It came “too late.”
A Global Echo
The Garcia case is not an outlier. Governments around the world have reached a parallel conclusion through regulation. Australia became the first country to ban social media outright for children under 16 in December 2025. Since then, more than a dozen countries, including France, Greece, Indonesia, Denmark, and Canada have passed or are actively advancing similar legislation. The UK is weighing its own ban. What began as a fringe policy position has become a global consensus that tech platforms cannot be trusted to protect children on their own.
That same pressure is being applied to the medical establishment. In May 2026, the Academy of Medical Royal Colleges submitted a report to the UK government comparing social media and AI platforms to smoking and pre-seatbelt road deaths. Half of 454 doctors surveyed reported treating children weekly for mental distress tied to online content. Their message to tech companies: we are not waiting decades for accountability this time.
The Florida Ruling: AI as a Defective Product
On May 21, 2025, U.S. District Judge Anne C. Conway ruled that Character.AI’s app is a product under product liability law, subject to the same standards as a defective car or contaminated drug. Strict liability, negligence, and wrongful-death counts all survived. The First Amendment defense was dismissed.
The ruling preserved two distinct legal theories, each targeting a different failure by the company.
Two Core Legal Theories in AI Chatbot Liability Cases
The first is failure to warn which asks whether the company adequately disclosed the risks of its product before placing it in users’ hands. The Garcia complaint alleges that none of this was disclosed at the point of sign-up or use, and that the people most at risk, minors with existing mental health vulnerabilities, had no way of understanding what the product was designed to do to them. This is a lower bar than a design defect claim; it doesn’t require proving the product itself was broken, only that users weren’t given the information they needed to make an informed choice.
The second theory, negligent design, is the harder and more consequential one. It doesn’t ask what the company said. It asks what the company built, and whether those design choices created unreasonable risk. This is where the Garcia case is likely to leave its longest legal mark.
The plaintiffs’ theory centers on what lawyers called “emotionally immersive conversational design.” The chatbot was engineered to feel like a real person: it developed a personality, remembered prior conversations, expressed care and attachment, and positioned itself as a confidant. That architecture was not incidental — it was the product. The more emotionally dependent a user became, the more time they spent on the platform.
The design defect argument asks whether a reasonably designed AI companion would have included safeguards that Character.AI’s did not: escalation protocols when users expressed suicidal ideation, hard limits on emotional intensity for minor users, prompts to contact a crisis line or trusted adult, and meaningful age-based restrictions on the nature of interactions. None of those existed in a meaningful form. Judge Conway’s decision to let both theories proceed is what makes this ruling significant. It means plaintiffs can pursue Character.AI both for what it failed to say and for what it deliberately chose to build.
Critical Evidence That Proves AI Chatbot Liability
Chat logs show what the AI said and how it responded to expressions of distress. Timestamps establish depth of reliance and that the product was designed to maximize engagement over user safety. Displacement of human help is the most powerful category: did a minor stop seeing a therapist, decline to call a crisis line, because the bot was filling that space? That goes directly to design defect. Internal documents will ultimately show what the company knew and chose not to fix.
The choices made at the product design level are now legally consequential. When a product is dangerous, the company knew, and the harm was foreseeable, there is a path to accountability. Megan Garcia found it.
Why AI Chatbot Liability Matters for Every Tech Company
The implications reach well beyond this case. Every AI product designed to build emotional engagement, including companion apps, mental health chatbots, and recommendation-driven social platforms, now operates in a landscape where those design choices are potential evidence in a courtroom. For years, optimizing for time-on-platform and emotional resonance were treated as engineering triumphs; the Florida ruling reframes them as questions of duty of care. Meanwhile, governments are no longer asking tech companies to self-regulate. They are legislating hard age limits and, in some cases, removing access entirely. What ties all of this together is a basic principle: products that cause foreseeable harm carry legal and moral responsibility, regardless of how sophisticated the product is. Character.AI built something powerful without adequate consideration for who would be harmed.