Three-tier AI red-teaming platform (PyRIT) testing this portfolio's own LLM/agent systems — real obfuscation attacks, Microsoft's Crescendo multi-turn technique, and agent manipulation testing, with vulnerabilities found and defense-in-depth verified against live infrastructure.
python ai-safety pyrit adversarial-ml prompt-injection anthropic llm-security owasp-llm-top-10 ai-red-teaming multi-agent-security
-
Updated
Sep 7, 2026 - Python