[ajmurmann]

I understood the bill to explicitly not target misuse of the AI (from the article: "Odd that ‘a model autonomously engaging in a sustained sequence of unsafe behavior’ only counts as an ‘AI safety incident’ if it is not ‘at the request of a user.’ If a user requests that, aren’t you supposed to ensure the model doesn’t do it? Sounds to me like a safety incident."). This seems to be entirely targeted at potential risk from a rogue AI. What regulation would you propose to address that risk?