We develop and advocate for pragmatic policies to reduce risks of severe harm from advanced AI.