Audn automates external black-box penetration testing for AI companies and enterprises. Existing cyber models are trained mostly on open-source data, while the highest-risk vulnerabilities now sit in closed-source production systems like binaries, public endpoints, network misconfigurations, AI behaviours. We have early pentester usage and a paid pilot with a YC S25 voice AI startup that generated revenue tied to a retailer contract.

LLM Infrastructure for adversarial, high-risk scenarios. Our adversarial AI helps authorised testers to simulate red teaming drills.