We have already seen this causing real problems for real people. Please see Jeremy Howard's lecture on ethics in AI for some real world examples.