OWASP AI Testing Guide security control for AI applications, focused on testing model robustness against adversarial inputs and perturbations designed to induce anomalous behavior or incorrect predictions.