On 11 May 2026, Google's Threat Intelligence Group confirmed a first: a criminal actor used an AI chatbot to write a working zero-day exploit. The exploit was a Python bypass of two-step login, intercepted before mass use. The report also named four state-linked clusters using AI tools in their operations.
AI wrote a functional attack from scratch for the first time. The incident has a named actor, a named file type, and a confirmed interception date.
