The voluntary AI safety commitments in the United States show that governments are paying attention. The difficult work will be turning principles into practice.