The lawsuit filed by the parents of Bowen Wan presents a deceptively simple question that has haunted AI policy circles for years: when a chatbot gives advice that kills someone, who is responsible?
Wan, who was taking anti-seizure medication, reportedly asked ChatGPT for guidance on reducing her dosage. The chatbot obliged with a tapering schedule. She followed it. She had a seizure. She died. Her parents now argue that OpenAI's product functioned as a de facto medical advisor without the safeguards, disclaimers, or liability frameworks that govern actual medical professionals.
The liability vacuum
AI companies have operated for years in a legal gray zone regarding the advice their systems dispense. OpenAI's terms of service contain boilerplate warnings against relying on ChatGPT for medical, legal, or financial decisions. But courts have historically been skeptical of fine-print disclaimers when products are marketed for general knowledge assistance and then used exactly as one might expect. The Wan case will test whether those disclaimers hold when the stakes are life and death.
The pharmaceutical industry learned decades ago that consumer-facing drug information requires extraordinary care. AI companies have absorbed none of those lessons. ChatGPT lacks the ability to verify a user's medical history, contraindications, or the specific formulation of their medication. It cannot follow up. It cannot refer to a specialist. It simply generates plausible-sounding text.
The regulatory silence
Neither the FDA nor the FTC has established clear frameworks for AI systems that provide health-adjacent information. The European Union's AI Act classifies medical devices as high-risk, but a general-purpose chatbot that occasionally answers medical questions exists in a different category entirely. This regulatory ambiguity has allowed AI companies to enjoy the commercial benefits of being useful for everything while bearing legal responsibility for almost nothing.
OpenAI has not publicly commented on the lawsuit. The company has previously pointed to its usage policies and the inherent limitations of large language models. But those arguments become harder to sustain when the product's interface encourages conversational, trusting interactions that mimic human expertise.
The precedent question
If the Wan family prevails, the implications extend far beyond OpenAI. Every AI company offering general-purpose assistants would face potential liability for advice given across domains where bad guidance can cause harm—medication, legal rights, financial decisions, even relationship counseling that leads to documented psychological damage. The industry's current posture assumes that users understand they are talking to a statistical model, not an expert. That assumption may not survive contact with a sympathetic jury.
Our take
The tragedy here is not that AI gave bad advice—humans do that constantly. The tragedy is that a multi-billion-dollar industry built products designed to feel authoritative while accepting zero responsibility for that authority. OpenAI wanted ChatGPT to be your research assistant, your tutor, your helpful companion. It got all of those things. Now it may also get the liability that comes with them. That would not be a bug in the legal system. It would be the system working exactly as intended.
