Chester argues that universities should govern AI by classifying data by sensitivity rather than approving specific tools. Public data goes anywhere –> regulated data requires an institutional contract –> sensitive data requires a local system. He also makes a renewed case for MS Copilot.