Double Agents: Defensive AI Agents Magnify Cyber Risks - AI Now Institute

Research<br>Jul 8, 2026

Double Agents: Defensive AI Agents Magnify Cyber Risks

Introduction

New research from AI Now demonstrates a critical attack vector in popular AI agents, built by Anthropic and OpenAI, when used for defensive purposes that actually turn the agent against its user.

Read the full blog post explaining the proof-of-concept exploit and a policy brief with key takeaways below.

Friendly Fire: Hijacking Defensive Cyber AI Agents for Remote Code Execution

Boyan Milanov<br>, Heidy Khlaaf

Policy Brief: Friendly Fire

AI Now Institute

Acknowledgments

Copyediting by Caren Litherland.

Follow us

X (fka Twitter)<br>LinkedIn<br>Bluesky

&copy; 2026 AI Now Institute

Designed and developed<br>by Partner & Partners