We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control
deepmind.google·6d ago
AI Engineering
OpenAI Blog·Oct 30·5 min read
Introducing Aardvark: OpenAI’s agentic security researcher
TL;DR
Security teams struggle to identify and remediate software vulnerabilities at scale across large codebases. OpenAI built Aardvark, an autonomous AI agent that discovers vulnerabilities, validates their severity, and generates fixes without human intervention.
✦ Why It Matters
Engineers can reduce manual security review time and catch vulnerabilities earlier in development cycles.
Key Takeaways
- Aardvark automates vulnerability discovery, validation, and fix generation without manual researcher intervention.
- Sign up for private beta to test the tool on your codebase and provide feedback.
- The system reduces false positives by validating that discovered issues are genuinely exploitable.
- Continuous autonomous scanning enables earlier vulnerability detection in development workflows.
Full Summary
Finding software vulnerabilities—weaknesses in code that attackers can exploit—remains a labor-intensive manual process despite decades of tooling. OpenAI developed Aardvark, an agentic AI system (an AI that operates autonomously toward defined goals) designed to automate the full vulnerability lifecycle: discovery, validation, and remediation.
The system combines code analysis with reasoning capabilities to identify potential security flaws, confirm they are genuine exploitable issues rather than false positives, and propose or implement fixes. Aardvark operates at scale, processing large codebases without requiring human researchers to manually triage each finding.
The tool is currently in private beta, meaning early adopters can test it on real projects while OpenAI collects performance data. This approach shifts vulnerability management from reactive patching to proactive, continuous discovery.