More than 1,500 AI projects are now vulnerable to a silent exploit

According to the latest research by ARIMLABS[.]AI, a critical security vulnerability (CVE-2025-47241) has been discovered in the widely used Browser Use framework — a dependency leveraged by more than 1,500 AI projects.

The issue enables zero-click agent hijacking, meaning an attacker can take control of an LLM-powered browsing agent simply by getting it to visit a malicious page — no user interaction required.

This raises serious concerns about the current state of security in autonomous AI agents, especially those that interact with the web.