Over 1,500 AI Projects Vulnerable to Silent Exploit

According to the latest research from ARIMLABS[.]AI, a critical vulnerability (CVE-2025-47241) has been discovered in the Browser Use framework, used by more than 1,500 AI projects. This flaw allows for zero-click agent hijacking, enabling an attacker to take control of a navigation agent powered by LLM simply by directing it to a malicious page, without any user interaction. This raises serious concerns about the current security of autonomous AI agents, particularly those that interact with the web.