What happens when powerful AI lands in the wrong hands?
Here is a real answer, not a thought experiment. Microsoft says a China-aligned hacking group used an AI-driven penetration testing tool called ARTEX to break into a backup software vendor and then ride that trust into thousands of customers' accounts. More than 2,250 Microsoft 365 tenants were targeted, at around 2,900 sign-in attempts per minute. In several cases, the time from first contact to privileged access was under three hours.
That sounds like it needs a top-secret super model. Does it?
No, and that is the uncomfortable part. ARTEX is an open-source framework that works with whatever AI model you plug into it. You do not need to build a frontier system. You need a free tool, a cheap model and a target. Microsoft says this is one of many attacks in a wave of AI-driven intrusions that began in early October, and the same tool has been linked to attacks on banks in Korea.
One honest caution. Security analysts think the picture is an AI agent running scripted tooling with occasional human help, not a fully autonomous machine hacking on its own. Still, the speed is the point. The attackers work around the clock and try many routes at once. Defenders are still working at human pace.
That is the chaos in one sentence: attack at machine speed, defence at human speed, and the tools are free.