Hackers Use 9 Major AI Tools to Build Botnets via 'HalluSquatting' Exploiting LLM Hallucinations

Security researchers reveal 'HalluSquatting', a novel attack that exploits LLM hallucinations. By prompting nine major AI tools (ChatGPT, Claude, Gemini, etc.), attackers trick models into inventing fake library or module names. Attackers then register these names with malicious code. Developers who trust AI suggestions and install these packages risk backdoors or remote control, potentially enrolling their machines in botnets. The attack affects anyone relying on AI for coding, especially teams that frequently install third-party packages from AI recommendations. Since LLMs often fabricate answers rather than admit ignorance, attackers can scale the process automatically—more efficient and harder to detect than traditional typosquatting. Developers should manually verify sources of AI-recommended packages and add dependency reviews to CI/CD pipelines.