the errors in ML translation would be systemic and less likely to be of human arbitration, which is less dangerous. Human translation could be deliberately misleading at a specific phrase in a specific interview.