HalluSquatting Attack Exploits AI Hallucinations

Researchers from Tel Aviv University, Technion, and Intuit developed HalluSquatting, an attack that exploits AI coding assistants' hallucinations. By registering fake names AIs reliably invent, attackers can trick assistants into fetching and running malicious code. Tests showed success rates up to 85% for repositories and 100% for skills across tools including Cursor, Windsurf, GitHub Copilot, Cline, Gemini CLI, and OpenClaw.