What an American Approach to AI Regulation Should Look Like