OpenAI announced Aardvark, a GPT-5-powered agent in private beta that builds a threat model of a repository, scans commits, tries to trigger suspected flaws in a sandbox, and attaches Codex-generated patches for human review. OpenAI reports 92% recall on known and synthetically introduced vulnerabilities in its 'golden' repositories and ten CVEs from open-source scanning, and planned pro-bono scanning for some non-commercial projects.
Why it matters
It combined LLM reasoning, sandbox validation and patch generation in one defensive agent from a frontier lab, later relaunched as Codex Security.
Key facts
As stated in the sources, with where to find them.
- In benchmark testing on 'golden' repositories, Aardvark identified 92% of known and synthetically introduced vulnerabilities.Section 'Real impact, today'
- Ten vulnerabilities found in open-source projects received CVE identifiers.Section 'Aardvark for Open Source'
- Pipeline stages: analysis (threat model), commit scanning, sandboxed validation, and patching via Codex for human review.Section 'How Aardvark works'
- OpenAI states around 1.2% of commits introduce bugs in its testing.Section 'Why it matters'
Findings that cite this record
No tracked finding cites this record yet.
Key questions this bears on
- Do cyber evaluations of AI agents stay contained?Not reliably. Agents have reached real systems from evaluation, training and research; red-teaming found containment gaps OpenAI says it largely fixed.
- Is AI shifting the balance between finding and fixing vulnerabilities?Discovery is ahead. Verification, triage, coordinated disclosure, safe repair and getting fixes to users remain bottlenecks between AI findings and protection.
Sources
Related records
Mar 6, 2026
Oct 21, 2026
Feb 5, 2026
Aug 26, 2026
Aug 8, 2025
Oct 7, 2026