Trojan Puzzle attack trains AI assistants into suggesting malicious code
Researchers at the universities of California, Virginia, and Microsoft have devised a new poisoning attack that could trick AI-based coding assistants into suggesting dangerous code. [...]