An Exploration of Attacks on AI Agent Systems

Summary

This paper investigates how prompt injection attacks can exploit both direct user inputs and inter-agent communication to leak sensitive information and trigger cross-agent infections. Its tests mainly highlight vulnerabilities in single-agent systems, and the authors extend the attacks to multi-agent frameworks to examine how weaknesses propagate across agents. The findings stress the need for more robust defenses as AI agents spread into real-world workflows.

Topics

• AI agents

Related items

• MediumCVE-2026-108759: mistral.rs sandbox escape via symlinks in mistralrs-code-execSimilar attack · NVD/CVE Database
• MediumCVE-2026-108756: Abilityai Trinity missing authorization in Telegram router binding operationsSimilar attack · NVD/CVE Database
• MediumCVE-2026-108600: open-multi-agent link following flaw in file_write tool sandboxSimilar attack · NVD/CVE Database
• MediumCVE-2026-108599: phi symlink escape in permission gate allows writes outside workspaceSimilar attack · NVD/CVE Database
• MediumCVE-2026-108592: mini-swe-agent information exposure in BubblewrapEnvironment sandboxSimilar attack · NVD/CVE Database