Adversarial testing techniques applied to generative artificial intelligence systems to identify vulnerabilities, misaligned behaviors, prompt injection, jailbreak, data leakage, and emerging security risks in language and multimodal models.