OpenAI Introduces Aardvark, an AI Agent for Security Research

OpenAI has introduced Aardvark, an autonomous security researcher powered by GPT-5, now available in private beta according to a company announcement. The agent is designed to help developers and security teams identify and fix vulnerabilities across large codebases.

Aardvark continuously monitors repositories to detect potential vulnerabilities, assess their exploitability, and propose targeted patches. It operates through a multi-stage process that includes repository analysis, commit scanning, validation of exploits in a sandboxed environment, and patch generation using OpenAI Codex. Each finding is accompanied by an annotated explanation and a Codex-generated fix for human review.

The system integrates with GitHub and existing development workflows, enabling continuous protection without interrupting engineering processes. In internal and partner testing, Aardvark identified 92% of known and synthetic vulnerabilities. It has also been applied to open-source projects, where ten findings received CVE identifiers.

Aardvark has been deployed internally at OpenAI for several months and has contributed to strengthening the company’s security posture. The private beta invites select partners to test the system’s performance across diverse environments before broader release.