China-Linked Hackers Use AI to Breach Hundreds Globally

The Speed Is the Weapon

Microsoft’s disclosure on October 21 does not read like a routine attribution bulletin. It reads like an alarm bell. The company revealed that a group it calls Storm-0940 — linked to the Chinese government — used an AI-powered penetration-testing tool named ARTEX to launch a global campaign against hundreds of organizations across nearly every major industry sector.

What makes this operation stand out is not just the scale but the velocity. In some cases, attackers went from identifying a target to obtaining privileged cloud access in under three hours. That timeline is short enough to turn what was once a methodical, weeks-long intrusion into something more akin to a raid.

The implications reach far beyond any single compromised organization. This is a glimpse into how artificial intelligence is quietly rewriting the rules of state-sponsored cyber warfare.

A Toolkit Turned Against Its Creators

ARTEX is an open-source tool. That detail matters. It was built for legitimate penetration testing — the kind of authorized security assessments that companies commission to find weaknesses before malicious actors do. ARTEX uses multiple large language model agents to automate the process: gathering information about a target, probing for vulnerabilities, executing attack techniques, and in some configurations, even extracting data.

The tool requires minimal human interaction. Set it running, and it can move through an attack chain on its own.

Storm-0940 weaponized that automation. Microsoft’s analysis indicates the group ran ARTEX as part of a sustained campaign that began in September 2026. The targets spanned aerospace, defense, automotive, chemical, energy, finance, government agencies, law firms, semiconductor manufacturers, and news organizations. The breadth suggests strategic intelligence collection rather than opportunistic crime.

The Numbers Behind the Noise

The raw data Microsoft published tells a story of aggressive, automated credential attacks.

Between October 10 and October 14, Storm-0940 targeted 327 accounts across eight organizations, generating more than 151,000 failed login attempts. On October 11 alone, the group recorded up to 2,900 failed logins in a single minute — over 111,000 in one day. The technique was password spraying: rapid-fire attempts to log into multiple accounts using common or leaked passwords.

The volume is impressive, but the actual compromise rate tells the more disturbing part. Microsoft confirmed that several organizations lost privileged cloud access. The group established persistent footholds across multiple environments, giving it repeated entry points long after the initial attack faded from logs.

The ConnectWise Pathway

A second vector in the campaign exploited ConnectWise, an American IT services company that provides managed IT management and cloud backup solutions to businesses. Attackers obtained authentication credentials for ConnectWise applications and used them to access customer environments — effectively turning a service provider into a doorway.

Microsoft said the attackers used ARTEX to attempt sign-ins to more than 2,250 customer Microsoft 365 environments through ConnectWise’s platform. Because the attackers had pre-granted permissions built into the service relationship, they could connect to customer systems without needing to impersonate employees through traditional phishing or interactive login flows. They walked through a back door that was already open.

ConnectWise responded on October 23, saying it expected to complete recovery environment testing and safety verification by the end of that weekend, with backup data restoration resuming early the following week if everything checked out. The company noted that individual customers would still need to verify whether their own environments had been accessed.

Who Wins and Who Loses

The winners in this scenario are transparent: any organization that relies on third-party IT management vendors, any company whose employees reuse passwords, and any entity that has not treated supply-chain credentials as a critical attack surface. The attackers did not need to breach firewalls. They needed a login key — and those are easier to obtain than ever.

The losers are the hundreds of organizations across multiple continents that found themselves on the receiving end of an automated, AI-assisted intrusion campaign. For government agencies and defense contractors, the damage may be measured in lost secrets. For semiconductor firms and aerospace companies, it could mean compromised intellectual property. For news organizations, it raises questions about source protection and editorial independence.

Why This Matters Beyond the Headlines

The geopolitical signal here is difficult to miss. Microsoft has publicly attributed the campaign to a group connected to the Chinese government. Beijing has remained silent — a pattern that is itself informative. China has a history of denying links to state-aligned cyber operations while quietly benefiting from the intelligence they produce.

What is less commonly discussed outside cybersecurity circles is what this episode says about the future of espionage. The barrier to launching sophisticated, multi-target attacks has never been lower. An open-source tool, a few cloud credentials, and an AI agent doing the heavy lifting — that is all Storm-0940 needed.

Western governments are unlikely to treat this as a routine incident. The combination of AI automation, state attribution, and cross-sector targeting creates political pressure for a response. Expect tighter regulations on AI tool usage, increased scrutiny of third-party service providers, and potentially new sanctions or export controls targeting the software infrastructure that makes campaigns like this possible.

What Comes Next

Organizations should assume their supply chain credentials are already in adversary hands until proven otherwise. The ConnectWise compromise demonstrates that the enemy is no longer knocking at the front door — it is walking through a side entrance that was left unlocked.

The broader lesson is that AI does not create new vulnerabilities so much as it accelerates the exploitation of existing ones. Password reuse, insufficient access controls, and overreliance on third-party vendors are not new problems. But they are now solvable problems at machine speed.

Storm-0940 has been operating since at least 2021. This latest campaign is only the most visible chapter. The real question is not whether another group will pick up a similar toolkit — the tool is already open source — but whether governments and enterprises will move fast enough to close the gaps that make AI-assisted intrusion this easy.