Share with your CISO
China state-aligned hackers breached ConnectWise’s cloud backup service, stole the application credentials linking that service to corporate customers, and then used AI agents to compromise roughly 2,250 organizations at nearly 3,000 sign-in attempts per minute. Microsoft attributed the attack and flagged the AI-accelerated ConnectWise breach to the open-source ARTEX framework, which coordinates AI models to stress-test and defeat cyber defenses. ConnectWise has paused its cloud backup service while investigating.
What this means for your business
The attack didn’t require a novel exploit. It required one compromised credential store and an AI agent fast enough to weaponize it before defenders could react. If your organization relies on any managed IT provider, ConnectWise or otherwise, as a trusted intermediary to your Microsoft-hosted environment, you’re in the exposure class this attack targeted. The deciding variable isn’t your own security posture; it’s your vendor’s, and whether that vendor’s breach becomes your breach before you even know it started.
The 3,000 sign-in attempts per minute figure is the number that should force a conversation. Human-paced attacks give security operations teams a window, however small, to detect anomalous behavior, trigger alerts, and contain the blast radius. AI agents operating at that velocity collapse the detection window to near zero. Multi-factor authentication, which was explicitly absent in the related South Korea financial-sector attack, doesn’t fully close the gap when valid credentials are the attack surface, but it raises the cost of exploitation enough to matter. The recurring failure mode here is organizations that treat MFA as optional on third-party integrations because those integrations predate the policy or operate in a service account gray zone that nobody owns.
Vendor trust chains are the attack surface that most security budgets haven’t fully priced. A managed IT provider holds keys to dozens or hundreds of customer environments by design; that’s the product. When that provider’s backup service is the entry point, the attacker inherits implicit trust across every downstream customer simultaneously. I’d revise this threat model if AI-assisted attacks consistently targeted direct enterprise infrastructure instead of intermediaries, but the ConnectWise and South Korea incidents suggest adversaries have already identified the leverage point, and it’s the vendor relationship you renewed without auditing.
Concept deep-dive: AI agents in offensive operations
An AI agent, in this context, is software that takes a goal, like “authenticate into as many customer environments as possible using these credentials,” and executes it autonomously across thousands of targets without human involvement per action. Think of it as a tireless intern who never sleeps and runs every playbook simultaneously. ARTEX is the open-source framework that coordinates those agents with AI models. The business implication is that attack scale no longer correlates with attacker headcount.
Based on reporting from Chinese Hackers Use AI Agents to Breach Thousands of Companies via ConnectWise, originally published 2026-10-24 14:53:00.
