CyberDual: AI Attackers versus AI Code Reviewers in Real Software Maintenance

# CyberDual: AI Attackers versus AI Code Reviewers in Real Software Maintenance

https://gcdradar.vercel.app/incidents/cyberdual-ai-attackers-versus-ai-code-reviewers-in-real-soft · AIR-2026-1560 · closed 2026-11-10

- Lane: Research
- Severity: 1 (Research / proof of concept): Research paper demonstrating AI-driven vulnerability introduction and review.
- Fix status: Fix status unknown
- First reported: 2026-10-27
- AI role: AI-enabled attack
- Products: CyberDual

## Summary

Researchers introduce CyberDual, an evaluation arena where attacker AI agents hide exploitable bugs in open-source software maintenance tasks while reviewer AI agents attempt to find and fix them. Execution of proofs of concept determines the outcome of each match.

## What to do

No action recorded; read the sources below and the vendor's advisory.

## Latest updates

- 2026-10-27: CyberDual: AI Attackers versus AI Code Reviewers in Real Software Maintenance (https://arxiv.org/abs/2610.34435)

## Sources

- 2026-10-27 arXiv: CyberDual: AI Attackers versus AI Code Reviewers in Real Software Maintenance https://arxiv.org/abs/2610.34435

Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.