Defending against dangerous AI should be a national priority — and we need more capable AI to do it

A rogue state or terrorist organization has little reason to observe an AI slowdown.