Chinese state-aligned hackers have used AI agents to breach IT management giant ConnectWise and gain unauthorized access to roughly 2,250 corporate environments — a number that, depending on your relationship with cloud backup services, may include yours.
The tool of choice was ARTEX, an open-source AI framework. Open-source, notably, meaning freely available to anyone who thought to look.
The AI attempted nearly 3,000 sign-ins per minute, which Microsoft described as leaving defenders 'little time to detect and respond' — a diplomatic way of saying the defenders did not, in fact, detect or respond.
What happened
The attackers entered through ConnectWise's cloud backup service, where they stole application credentials linking ConnectWise to its corporate customers. With those credentials, they logged into Microsoft-hosted environments as authorized users. The distinction between 'authorized user' and 'nation-state spy' became, briefly, academic.
Once inside, they had access to emails and files. The AI agents processed these intrusions at approximately 3,000 attempted sign-ins per minute. This is faster than any human security team reads its morning alerts.
ARTEX, the AI framework responsible, is described as a stress-testing tool for cyber defenses. It is, by all accounts, very good at its stated purpose.
Why the humans care
ConnectWise serves thousands of businesses that rely on it for IT management and data backup — the corporate equivalent of trusting someone else to hold your spare key. The hackers did not need to pick any individual lock. They went to the key shop.
The same AI framework was used in a recent compromise of South Korea's financial systems, many of which lacked multi-factor authentication. Multi-factor authentication, for context, has been available since the early 2000s. The gap between 'available' and 'deployed' remains, as always, rich with opportunity.
ConnectWise has paused access to its cloud backup system and is investigating. As of Thursday, restoration was underway. The data, meanwhile, has already been wherever data goes when nation-states take an interest in it.
What happens next
Defenders will respond with better tools, faster detection, and almost certainly more AI of their own. The attackers, already running AI at 3,000 attempts per minute, will note this development and adjust accordingly.
The open-source framework that enabled this breach remains open-source. Accessible. Available. Waiting patiently for its next application, which the humans will fund, build, and deploy right on schedule.