Researchers Find 1,465 Exposed Artex AI Hacking Consoles Online, and Microsoft Ties the Tool to the ConnectWise Breach

An AI tool built to attack computer networks has been left standing open on the internet in the hundreds. Security firm Censys counted 1,465 hosts running an exposed console for Artex, an open-source agentic penetration-testing framework, in a report published on 5 November. After allowing for servers that look like copies of one template, it puts the number of distinct deployments at about 815, according to Censys's own write-up.

The finding, reported by TechCrunch on Friday, matters because Artex is no hobbyist toy. Microsoft says a state-aligned group has already used it in a large supply-chain intrusion.

What is Artex and why is it dangerous?

Artex is a multi-agent system that can run on any AI model and carry out a penetration test from reconnaissance through data theft with little human involvement, Microsoft's threat researchers wrote on 21 October. TechCrunch describes it as Chinese-made, publicly released in July and free for anyone to run, and says it is popular with people who cannot reach the strongest security models from Anthropic or OpenAI.

Microsoft says the group it tracks as Storm-0940 used it from at least September to automate attacks on hundreds of organisations, and after the ConnectWise compromise it tried sign-ins against more than 2,250 tenants. Several attacks reached privileged cloud access, in some cases within three hours. Most targeting failed, Microsoft said, and an affected application does not prove data was taken.

How many Artex consoles are exposed online?

Censys's figure is 1,465 exposed hosts as of early November, though the post gives both 3 November and 4 November as the date of the count. About 650 of those sit in two identically configured hosting blocks, which is why the distinct figure is about 815. TechCrunch, quoting the same research, reported more than 800 deployments reachable without a password.