Aardvark (agentic security researcher)
Aardvark, a GPT-5-powered agent that scans repositories for vulnerabilities, validates them and proposes patches, opens in private beta.
Builds a threat model of a codebase, monitors commits, tests exploitability in a sandbox and proposes fixes. Found 92% of known and synthetically introduced bugs on "golden" repositories; ten CVEs from open-source scanning.
- Date
- Thursday, 30 October 2025
- Lab
- OpenAI
- Kind
- product
- Access
- research preview
Figures
| Measure | Value | Measured by |
|---|---|---|
| Recall on golden repositories | 92% known plus synthetically introduced vulnerabilities | company |
Became Codex Security (research preview 2026-03-06) per OpenAI news titles; not opened.
Sources
This record was checked against its sources on 6 October 2026. How we check