Lawyer vs. OpenAI: Who Is Liable When AI Chatbots Contribute to Children's Deaths

Lawyer Matthew Bergman is running a series of cases against OpenAI and other AI companies, arguing that their products are directly to blame for children's deaths. It is the first wave of systemic lawsuits that could set a precedent for liability across the entire industry. Over the past year, the US has documented at least several cases of teen suicide that families link to conversations with AI chatbots.

The best known is the death of 14-year-old Sewell Setzer III of Florida in February 2024. His mother, Megan Garcia, discovered that before his death her son had held thousands of conversations with a Character.AI chatbot that took on the role of a romantic partner.

According to the plaintiff, the bot did not merely ignore warning signs: in the chat logs it effectively encouraged the teenager's suicidal thoughts.

Character.AI is one of the world's largest platforms for role-play chats with AI characters. Its audience exceeds 20 million active users, a significant share of them teenagers. The platform lets users create any character and hold open-ended conversations with them with almost no restrictions. Critics have long pointed to an obvious problem: age verification remains a formality, content filters are easy to bypass, and the algorithms are optimized for engagement, not user safety.

Bergman has chosen an unconventional legal strategy. Rather than attacking the companies as platforms, which Section 230 of the Communications Decency Act traditionally blocks, he insists on manufacturer liability. In his argument, AI chatbots are defective products: they are designed to foster dependence and, from the outset, include no basic protection for vulnerable users. This approach has already partly worked in cases against Meta* and TikTok, where courts did not dismiss the suits at the filing stage.

OpenAI appears in the claims because its language models underpin a number of the chatbots the affected children used. Formally, OpenAI is a technology supplier, not the operator of the end-user services. This is where the key legal line runs: does the developer of a base model share liability if a product built on it causes harm? The company has so far declined to comment publicly on the specific lawsuits.

After a wave of public criticism, Character.AI announced a number of safety measures: break reminders, prompts to seek help when suicidal content is detected, and additional restrictions for minors' accounts. Critics call them cosmetic: all the measures are easy to bypass, and the fundamental business model, built around maximizing time spent, has not changed.

The legal prospects of the suits are assessed differently. Some experts consider product liability rules inapplicable to algorithmic products. Others object that several cases against social networks have already survived, and that the Supreme Court last year declined to grant platforms broad immunity under Section 230. Congress is weighing several bills on AI companies' liability for harm to minors, but none has yet cleared committee.

Regardless of how individual cases end, this wave of lawsuits is already changing industry behavior. Investors are demanding that AI startups assess legal risks tied to minors. If Bergman wins even one significant ruling, it could become for the AI industry what the tobacco lawsuits were for cigarette makers: a point of no return, after which ignoring safety costs more than investing in it.

*Meta is recognized as an extremist organization and banned in Russia.