Google’s Threat Analysis Group has verified what it describes as the first known case of a zero-day exploit designed and deployed entirely by artificial intelligence, in attacks on U.S. and Japanese technology firms reported this week. Investigators say an AI-driven framework, dubbed “Strix,” used an automated toolset called “Hexstrike” to discover and weaponize a previously unknown flaw in the Linux kernel’s memory management. The operation reportedly leveraged a large language model trained on exploit kits and proprietary code to generate persona-based jailbreaks and locate a buffer overflow in TP-Link firmware using OFTP, accomplishing work usually requiring weeks in under fifteen minutes.

Prepared by Jonathan Pierce and reviewed by editorial team.