With fantastic power arrives fantastic accountability. As a fresh generation of AI creators sets out to construct strong models, how can they make sure they are doing so responsibly? Robustness: AI-powered methods should be actively defended from adversarial assaults, minimizing security haza