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

Developers should certainly work to make AI safe. But our best protection against dangerous AI may ultimately be other, more capable AI. Consider ...