Red Teaming

Actively and systematically probing AI systems for weaknesses and vulnerabilities, especially those not considered during its design.