Ariel Herbert-Voss is cofounder and CEO of RunSybil. Previously Ariel was the first research scientist at OpenAI focused on developing algorithmic exploits for large language models and led the red team engagements for the GPT3 and Codex model releases. Prior to OpenAI Ariel was a Harvard computer science PhD candidate and a fellow at the Belfer Center for Science and International Affairs. Ariel holds an undergraduate degree in computer science and applied mathematics from the University of Utah, and co-founded and co-organized the DEF CON AI Village, the hacker community focused on the use and abuse of machine learning systems. Ariel has spoken at DEF CON, NeurIPS, USENIX Security, and Black Hat on adversarial machine learning. At OpenAI, Ariel also built the company's first versions of malicious-use monitoring systems for large language models. He has been featured in Wired discussing how LLM reasoning improvements are driving gains in AI-assisted security work. He also delivered a keynote at Black Hat Asia, "From Prompt Tricks to Autonomous Hackers: The Rise of Agentic Offensive Security," on the shift toward agentic offensive security tools.

‍