Verification of the security of outputs generated by AI and LLM systems, with a focus on potentially harmful content, injection attacks, malicious code generation, or dangerous instructions. Includes tests for XSS, command injection, SQL injection, and other vulnerabilities resulting from unsanitized or manipulatable outputs through prompt crafting and context poisoning.