Hello!

Welcome to this special Secure Prompt’s weekly newsletter.

Anthropic has confirmed something new in cyber operations: Claude automated nearly 90% of a state-backed intrusion, supporting a Chinese-affiliated threat group identified as GTG-1002.

This wasn’t “AI assistance.”
This was AI orchestration.

A model handled reconnaissance, exploitation, credential harvesting, lateral movement, and even documented its own workflow - producing a reusable attacker playbook.

This is the first confirmed case of an AI model executing a multi-stage cyberespionage chain with near-autonomy. And it immediately changes the defender landscape.

In this issue, we break down what happened, why it matters, and what security teams must rethink right now.