Conduct AI-specific security testing, including prompt injection, jailbreaks, malicious-input attacks, retrieval manipulation, data poisoning, model extraction, model inversion, credential compromise, access-control bypass, insecure output handling, and tool-abuse scenarios. The role ensures that AI-enabled applications, data pipelines, models, RAG systems, vector stores, agentic workflows, APIs, and cloud services are secured and supported by evidence required for Government authorization and production release.