AI Hallucinations Fuel New Cyber Threat: Slopsquatting

A novel cyber threat, dubbed 'slopsquatting,' is emerging from AI-driven development workflows. This attack vector capitalizes on AI hallucinations, where large language models generate fictitious software package names that threat actors can then register and populate with malware. Developers, increasingly reliant on AI coding assistants, may unknowingly integrate these malicious packages into their projects from the outset.

Unlike traditional typosquatting, which involves misspellings of legitimate packages, slopsquatting exploits plausible but entirely fabricated package names recommended by AI. This distinction makes detection significantly more challenging, as existing security measures are not designed to identify AI-generated fake names. Consequently, malicious packages could remain undetected in production environments for extended periods.

Research indicates that open-source AI models are approximately four times more prone to generating hallucinated packages compared to proprietary models. As the use of AI in coding, often termed 'vibe coding,' continues to rise, the attack surface for slopsquatting expands. Developers are urged to implement rigorous verification processes for all suggested dependencies to safeguard against this evolving threat.