A new policy framework from the Aspen Institute would establish clear processes for states to follow when AI produces unwanted or dangerous results.