OpenAI has launched a new artificial intelligence (AI) called “Aardvark” designed to detect vulnerabilities in software code. The main purpose of this system is to identify errors and weaknesses in code that could potentially be exploited by hackers. Built on the latest GPT-5 technology, Aardvark connects to code repositories where developers store their programs and begins analyzing the code completely autonomously.
When the system detects a potential vulnerability, it immediately verifies whether it is truly dangerous. To do this, the code is tested in a secure, isolated environment separate from the main program. Only after confirming that the issue is indeed risky does Aardvark generate a detailed report showing the severity level of each problem. At this stage, OpenAI’s other AI model, Codex, is activated to provide ready-made solutions and fixes for the identified vulnerabilities.
According to the company, Aardvark has successfully detected 92% of security vulnerabilities during testing, and OpenAI has been using the system for several months to ensure the safety of its internal software. It should be noted that Aardvark is not yet publicly available. The program is still in closed beta testing, meaning only a limited number of users who apply and receive approval can access it.
