I used to think of AI governance as distant from building. Now it feels directly connected to what kind of products people can trust.