Ariel Fogel

About

I’m an AI security researcher and product engineer who builds and secures production agentic systems end to end. At Pillar Security, I developed RedGraph end to end: an agentic red-teaming system that maps an AI system’s attack surface, performs threat modeling, and exploits attack paths enabled by LLM behavior and tool abuse. RedGraph turns ambiguous AI security risks into concrete, reproducible tests. My research focuses on prompt injection, agentic risk, tool misuse, data exposure, AI supply-chain vulnerabilities, and the trust boundaries of LLM applications. I’ve published AI security research, discovered vulnerabilities in coding-agent workflows, contributed to OWASP AI security initiatives, helped brief congressional staff on AI risk, and serve as a founding contributor, core team member, and spec workstream lead for the Agent Control Standard. I’m strongest in the zero-to-one phase, where the work requires deep research, clear success criteria, fast hypothesis validation, and systems built from first principles. I like working close to users, reading data carefully, and building AI systems with evaluation and testing as part of the development lifecycle, not as an afterthoug

Experience

Education