Red Teaming

A practice where skilled testers deliberately try to break an AI model and find its weaknesses before public release.