Rows of servers in a data center. Microsoft says most of the intrusions were stopped, but the attackers kept high-level access at many targets. (Image: BalticServers.com / Wikimedia Commons — CC BY-SA 3.0)
A hacking group linked to China used an artificial intelligence-driven penetration testing tool to compromise the IT services firm ConnectWise as part of a campaign that targeted hundreds of organizations around the world, Microsoft said on Wednesday.
According to Microsoft, the group it tracks as Storm-0940 carried out the campaign using ARTEX, an open-source penetration testing framework. Tools of this kind are built for the legitimate work of testing network security — but ARTEX lets users plug in any AI model to automate tasks including reconnaissance and data exfiltration, and can be configured to run autonomously.
The targeting was sweeping: organizations across aerospace, defense, automotive, chemicals, energy, financial services, government, legal, manufacturing, media, military, semiconductor, software, technology, think tanks and nonprofits were all in the campaign's sights, Microsoft said. Most of the attempts were not successful — but at many organizations, Storm-0940 achieved what Microsoft called "privileged cloud persistence," a high level of access that can survive even after cleanup efforts begin.
ConnectWise, whose software is used by IT service providers to manage their clients' systems, did not immediately respond to a request for comment on Microsoft's conclusions. In a statement posted on its website on October 16, the company said it was "investigating recent attempts of unauthorized access associated with the ConnectWise Cloud Backup environment." The Chinese Embassy in Washington did not immediately respond to a request for comment; Beijing regularly denies conducting or condoning cyber-enabled data theft.
The disclosure is the latest sign that AI is compressing the work of hacking campaigns as much as it is every other kind of knowledge work. Automating reconnaissance and exfiltration lets a small team probe hundreds of targets simultaneously — and defenders are left racing to patch the legitimate tools that attackers have learned to turn against their owners. For the hundreds of organizations in Storm-0940's sights, Microsoft's message is blunt: check whether the access the group won quietly has really been closed off.
Also read: Nvidia's DSX POD packs an entire AI factory into a container | Meta ordered to pay $25 billion in the New Mexico privacy case