On March 7, 2026, OpenAI released Codex Security, an AI-driven application security agent tool capable of automatically detecting security vulnerabilities in code, verifying suspected issues, and providing remediation solutions. Derived from the previously tested security research agent Aardvark, the tool is now available as a 'research preview' for Enterprise, Business, and Education tier customers, offering a one-month free trial. Codex Security conducts comprehensive analysis of code repositories, tests suspected vulnerabilities in isolated sandbox environments, generates exploit examples and remediation recommendations, helping enterprises shift security checks earlier into the development process. During the testing phase, the tool identified nearly 800 critical vulnerabilities and over 10,500 high-risk issues in publicly exposed code repositories, successfully locating vulnerabilities in open-source projects such as OpenSSH, GnuTLS, and Chromium.