botnet

Researchers at Tel Aviv University have chained two well-known AI weaknesses, hallucination and prompt injection, into a technique that can trick AI coding assistants into installing botnet malware on a user's machine.