#Microsoft's #SatyaNadella called for #AIsystems with #humancontrolled #emergencybrakes, #containment and #independentcontrols. He urged deterministic design and treating models as insider risks, citing observability principles including auditability and incident disclosure.…