What happened
OpenAI introduced Aardvark, a GPT-5 powered AI agent that autonomously scans, exploits, and patches software vulnerabilities. It operates by analyzing code behavior and running tests to identify and fix bugs without relying on traditional methods like fuzzing. Aardvark is currently in an invite-only beta and has been tested internally and with external partners.
Why it matters
- Aardvark automates the detection and remediation of software security flaws, potentially reducing human workload in cybersecurity.
- It can identify not only security vulnerabilities but also logic and privacy bugs in codebases.
- OpenAI plans to offer free scanning for noncommercial open source repositories, expanding access to automated security tools.
Key facts
- 1. Aardvark was released by OpenAI on Thursday, October 30, 2025, as part of a beta program.
- 2. The AI identified 92% of known and synthetic vulnerabilities in tested repositories.
- 3. The model uses GPT-5 technology and does not rely on traditional program analysis techniques.
