ZKTOR's developer places human review and appeals alongside protective AI, separating risk detection from decisions affecting participation.