OpenAI releases bug-finding intelligent agent Aardvark: fully automatic code reading, vulnerability detection, and fix writing

OpenAI has released Aardvark, a GPT-5-powered agent capable of automatically detecting and fixing security vulnerabilities in large-scale codebases. Aardvark can identify 92% of known and manually injected vulnerabilities and pinpoint issues under complex conditions. By monitoring code commits, analyzing and validating vulnerabilities, and generating repair suggestions, it integrates deeply with OpenAI Codex to produce patches for detected issues.