RunSybil

RunSybil builds an AI-native offensive security platform, centered on an autonomous agent called Sybil, that runs continuous automated penetration testing against live enterprise environments rather than the traditional point-in-time pentest engagement. Instead of a scheduled annual or quarterly test, Sybil runs continuously against every new deployment, re-evaluates only what changed, and validates findings through live exploitation, an approach the company says reduces false positives substantially compared with conventional vulnerability scanning, a vendor claim not yet independently benchmarked.

Founded in 2023 by former OpenAI security researcher Ari Herbert-Voss and former Meta offensive security lead Vlad Ionescu, and based in San Francisco, RunSybil raised a $40 million round led by Khosla Ventures, with participation from Menlo Ventures and angel investors from OpenAI, Palo Alto Networks, Stripe, and Google. That is a substantial raise for a two-year-old company and reflects strong investor conviction in AI-native offensive security as a category, though it is not itself evidence of product efficacy.

RunSybil states it serves select startups along with undisclosed financial institutions and Fortune 500 customers, but does not publicly name reference customers, and has not participated in independent evaluations such as MITRE ATT&CK Evaluations (which in any case are oriented toward detection products rather than offensive testing tools). Its claims of continuous, low-false-positive validated exploitation should be treated as vendor-reported pending independent, named case studies.