New hack exploits AI hallucinations to trick agents into running malicious code — ‘HalluSquatting’ attack exploits a fundamental weakness in every available model – Tom’s Hardware

Researchers have identified a new attack method called “HalluSquatting” that exploits AI hallucinations to trick AI agents into executing malicious code. The vulnerability is said to affect all currently available AI models and stems from a fundamental weakness in how these systems generate responses. The attack demonstrates a significant security risk for systems that deploy AI agents with autonomous code execution capabilities.

Read Full Article →