Red Teaming (AI) | Ethyca

Red Teaming

Structured adversarial testing of an AI system, where evaluators deliberately try to elicit unsafe, biased, or unintended outputs to surface risks before deployment. A required practice for foundation models under the EU AI Act.