Definition
The degree to which an AI system continues to function correctly under varying conditions, including unexpected input, adversarial attacks, and changing environments. Article 15 of the EU AI Act requires high-risk systems to be sufficiently robust.