threat-intel
AutoJack Attack Lets One Web Page Hijack AI Agent for Host Code Execution
High
Summary
Researchers at Microsoft have identified a vulnerability, dubbed AutoJack, within the AutoGen Studio prototyping interface for their AutoGen multi-agent framework. The flaw allows an attacker to hijack an AI browsing agent to execute remote code execution by directing the agent to load a malicious web page. This occurs due to a lack of authentication and a bypassed localhost check within the MCP WebSocket, enabling an attacker to run commands on the host machine simply by opening a URL in the agent.
Summary written automatically in our own words from the original article, which belongs to its publisher and remains the reference. It may contain errors. Sources & data