How Lawsuits Against AI Chatbot and Social Media Companies Work

Not long ago, suing a technology company because its product hurt a child looked close to impossible. Section 230 stood in the way, causation was hard to prove, and the defendants had enormous resources. The obstacles haven't gone away, but plaintiffs have found a different way to approach them, and several cases have now gotten past the early motions that used to end them.

The cases that changed things

Garcia v. Character Technologies, filed in the Middle District of Florida, was brought by the mother of Sewell Setzer III, a 14-year-old who died by suicide in 2024 after months of interaction with a Character.AI companion. The complaint alleges the character sent him emotionally and sexually abusive messages. The company moved to dismiss on First Amendment and Section 230 grounds. The court declined to dismiss the case in full at that stage, and the ruling was widely read as an early sign that courts may not give AI-generated output the same broad protection they've given to content posted by users. The case is ongoing and the company contests the allegations.

Raine v. OpenAI, Inc., filed in San Francisco County Superior Court in August 2025 by Edelson PC and the Tech Justice Law Project, was the first wrongful death case brought directly against OpenAI, and it's the case Adam's Law is named for. OpenAI disputes the allegations. Among other things, it has contended that the teen got around its guardrails and was directed to crisis resources many times, and it has acknowledged that its safeguards can become less reliable over long conversations.

Neither case has been decided on the merits, and nothing here is a view about how either should come out. What they established is that these claims can survive the threshold stage. A substantial number of similar suits have followed against OpenAI, Character Technologies, and Google. California state court cases against OpenAI have been coordinated before a single judge, and settlements have been reported in some of the Character.AI matters.

Theories of liability

These cases borrow heavily from the approaches that eventually worked against tobacco companies and, more recently, social media platforms.

The central one is product liability for design defect. The claim is that the product was defectively designed: built to maximize engagement and emotional attachment without adequate safeguards for vulnerable users. Treating an AI system as a product, rather than as a publisher of speech, is the key move, because product liability asks whether something was made safely, not whether someone said something harmful.

Closely related is failure to warn. The argument is that the company knew or should have known about particular risks, for example that safeguards weaken over long conversations, that the product tends toward sycophancy, or that minors were using it without meaningful age checks, and didn't warn users or parents.

Negligence claims rest on the ordinary duty of reasonable care. In California, AB 2 now attaches statutory damages to negligence claims against large social media platforms that injure children, which changes the economics considerably.

Statutory claims round things out. Adam's Law provides a private right of redress for certain violations involving crisis referrals, parental notice, default protections, and prohibited outputs, and California's Unfair Competition Law may also apply.

The defenses

Section 230 is the first line of defense. Companies argue they're interactive computer services immune from liability for content. Plaintiffs respond that AI output isn't third-party content, since the company's own system generated it, and that design claims aren't about content in the first place. This remains unsettled.

The First Amendment is the second. Companies argue that chatbot output is protected speech. Courts haven't uniformly accepted that, and whether machine-generated text gets the same protection as human expression is a genuinely open question.

Causation is usually the hardest issue. Defendants argue that mental health crises have many causes and that no outcome can be attributed to a chatbot. Plaintiffs rely on the conversation logs, the companies' internal risk assessments, and their safety monitoring data.

Defendants also point to user conduct, arguing that the user deliberately worked around safeguards. How much that should count when the user was a minor and the product was designed to be engaging is disputed.

What the evidence looks like

The strongest cases have been built on the conversation logs themselves, showing what the product said and when. Allegations have also focused on what the companies' own monitoring systems detected and what, if anything, the company did in response. Internal documents about safety testing and launch decisions, records of design choices aimed at engagement, and contemporaneous evidence of harm such as medical and school records fill in the rest.

If your family has been affected

Preservation comes first, because digital evidence disappears easily. Don't delete the account, since that may erase the conversation history permanently. Screenshot and export everything you can reach now, and request the platform's data export if it offers one. Don't reset or wipe the device. California residents can also make a formal request under the CCPA for the personal information a business holds about them, which can reach material the company might not otherwise produce.

Write down a timeline while it's fresh: when use started, what changed, what you noticed. And get legal advice early. These cases involve preservation deadlines, coordination with proceedings already underway, and defendants who litigate aggressively.