“Legal responsibility often turns on intent—traditionally understood as a fact about an actor’s state of mind. An AI system’s capacity for responsibility would seem to depend on its ability to form mental states. But intent can function not only as fact but as fiction—a construct ascribed to an actor by others. While law claims to prize intent as a fact about an actor’s mental state, liability often turns on the perceptions of those affected. Without resolving debates over whether software can possess mental states, AI systems may still bear legal responsibility through the intent attributed to them.”