I'm tired of how agentic language is used to describe AI models. This language avoids placing responsibility where it should be, with humans
When a plane crashes we don't think *the plane itself* is somehow morally responsible. It's safety regimes, pilots, etc. Same here