Safe Reasoning Over Ontologies