Model alignment shapes how AI behaves, but it cannot record who used which model, on what data, and under whose policy. Real governance lives around the model.