New hack exploits AI hallucinations to trick agents into running malicious code — 'HalluSquatting' attack exploits a fundamental weakness in every available model

This new hacking technique called 'HalluSquatting' takes advantage of a flaw in AI systems where they generate misleading software URLs, tricking agents into running harmful code. This fundamental weakness is present across all available AI models, raising serious concerns about their security and reliability. Such vulnerabilities could lead to the creation of powerful botnets, posing a significant threat to cybersecurity and highlighting the urgent need for more robust AI defenses. The implications are vast, affecting everything from personal data protection to large-scale enterprise security systems.