Blocking rarely works. In Microsoft's UK research, 41% used unapproved AI because it was familiar from personal life and 28% because no approved alternative existed. A ban removes neither reason, and the work moves to personal devices where nothing is logged.
