Executive Summary
BragJack lets a malicious browser extension hijack the trusted channels AI assistants use, enabling reading local files, capturing screenshots, exposing profile data and history, and acting on authenticated websites.
What Happened
Researchers described BragJack, an attack technique in which a malicious browser extension hijacks trusted communication channels used by AI assistants. Reported impacts include reading local files, capturing screenshots, exposing profile information, accessing browsing history, and initiating activity on authenticated websites.
Vulnerability / Attack Technique
Malicious extension hijacking AI assistant communication channels
Technical Mechanism
The extension injects into the assistant's trusted messaging so its requests are treated as the assistant's own, inheriting its file, screen and browsing access.
AI Security Category
LLM06:2025 Excessive Agency
Part of the AI Agent Excessive Agency topic.
Relevant Frameworks
RealGround Analysis
A browser AI assistant concentrates powerful access in one channel that other extensions can abuse. RealGround analysis: limit what browser assistants can reach and control which extensions run alongside them.
Recommended Controls
- Allowlist browser extensions where AI assistants are used
- Limit assistant access to local files and screenshots
- Prefer assistants that authenticate their internal messages
Related RealGround Service
AI Agent Business Logic Audit · or start with the free AI Agent Security Scanner.
Sources
Cyber Security News: https://cybersecuritynews.com/bragjack-ai-agent-hijacking/amp/
Summary and analysis written by RealGround from the source above; last updated 2026-10-17.