OpenAI Introduces Aardvark, an AI Security Agent Powered by GPT-5

OpenAI has announced the launch of Aardvark, an autonomous AI security agent powered by GPT-5 that aims to revolutionise how organisations discover and fix software vulnerabilities, with private beta currently available.

The system operates as an autonomous security researcher that continuously monitors source code repositories to identify vulnerabilities, assess their exploitability, and propose targeted patches, using large language model–powered reasoning to read code, analyse it, write tests, and employ various tools.

In benchmark testing on repositories with known vulnerabilities, Aardvark achieved a 92 percent detection rate, and it has contributed to open-source security by responsibly disclosing numerous vulnerabilities discovered in open-source projects, ten of which have received CVEs.

The company plans pro‑bono scanning for select non-commercial open-source repositories to strengthen the wider software supply chain and notes that there were over 40,000 CVEs reported in 2024, with about 1.2 percent of code commits introducing bugs. With updated coordinated disclosure policies and plans to broaden availability beyond private beta, according to OpenAI, Aardvark could broaden access to advanced security expertise.