Chinese hackers used AI agents to breach ConnectWise and spy on thousands

China state-aligned hackers breached the IT management giant ConnectWise, stealing application credentials that allowed them to spy on thousands of corporate customers. Security researchers at Microsoft confirmed the intrusion into ConnectWise’s cloud backup service, which exposed data links used by companies to back up their information.

The attackers leveraged artificial intelligence to automate the exploitation of these stolen credentials, targeting approximately 2,250 organizations in a rapid, coordinated campaign. This incident highlights a growing threat where AI tools are no longer just theoretical risks but active components in large-scale cyber espionage operations.

What happened

Security researchers at Microsoft identified that hackers broke into ConnectWise’s cloud backup service. The intruders did not just steal static data; they extracted application credentials. These credentials are critical because they establish the trusted data links between ConnectWise and its corporate customers. Many of these customers rely on the cloud service to back up their company’s data, meaning the compromised credentials provided a direct pathway into their internal systems.

Once the hackers possessed these credentials, they could log in to the customers’ Microsoft-hosted environments as if they were authorized users. This level of access allowed them to view emails, access files, and monitor activities without triggering immediate suspicion. The breach effectively turned a trusted vendor relationship into a vulnerability for thousands of businesses that had placed their confidence in ConnectWise’s infrastructure.

ConnectWise responded by pausing access to its cloud backup system to contain the damage. The company stated it is actively investigating the breach and working on restoring the service. As of Thursday, restoration efforts were underway, but the scope of the data exposure remains a significant concern for the affected organizations.

How it works

The scale and speed of this attack were driven by the use of AI agents. Microsoft reported that the hackers attempted sign-ins at a rate of almost 3,000 per minute. This volume is difficult for human security teams to manage manually, as it overwhelms traditional monitoring tools and leaves defenders with little time to detect and respond to the anomalies.

The specific tool used was ARTEX, an open-source AI framework designed to work with AI models to stress-test cyber defenses. In this context, the hackers repurposed a tool intended for defensive testing into an offensive weapon. ARTEX automated the process of using the stolen credentials to attempt access across thousands of accounts simultaneously, bypassing the need for manual intervention and allowing the attackers to scale their operations efficiently.

This same AI software was previously linked to a compromise of South Korea’s financial systems. In that instance, many targeted systems had weak security protocols and lacked multi-factor authentication, making them vulnerable to automated brute-force or credential-stuffing attacks. The recurrence of ARTEX in these high-profile breaches suggests that state-aligned actors are standardizing their toolkits around accessible, powerful AI frameworks.

Key details

• Target: ConnectWise, a major IT management and remote access provider.
• Attacker: China state-aligned hackers, according to Microsoft security researchers.
• Method: Breach of ConnectWise’s cloud backup service to steal application credentials.
• Scale: Approximately 2,250 organizations were targeted and accessed.
• Tool: ARTEX, an open-source AI framework used to automate sign-in attempts at nearly 3,000 per minute.
• Impact: Hackers gained authorized-level access to Microsoft-hosted environments, including emails and files.

Why it matters

For software developers and IT leads, this incident underscores the fragility of supply chain security. When a vendor like ConnectWise is compromised, the risk cascades to every customer who trusts that vendor with their data links. It is no longer sufficient to secure only your own perimeter; you must also evaluate the security posture of the third-party services that integrate with your infrastructure. The theft of application credentials is particularly dangerous because it bypasses user-level protections, granting attackers systemic access.

The use of AI agents like ARTEX changes the dynamics of incident response. Traditional security operations centers rely on alerts and human investigation, but an attack speed of 3,000 sign-ins per minute compresses the decision window to seconds. Defenders must now assume that automated threats will outpace manual responses. This necessitates a shift toward automated defense mechanisms that can match the speed of AI-driven attacks, such as real-time anomaly detection and automated credential revocation.

Furthermore, the reuse of ARTEX in different geopolitical contexts, from South Korea’s financial sector to global IT management firms, indicates that these tools are becoming commoditized among advanced threat actors. Organizations can no longer treat AI-enhanced cyberattacks as rare or experimental. They are a persistent reality that requires updated security architectures, including strict enforcement of multi-factor authentication and zero-trust principles for all service-to-service connections.

What you can do

• Audit all third-party integrations and identify which services hold application credentials for your environment.
• Implement strict multi-factor authentication for all administrative and service accounts, especially those linked to cloud backups.
• Monitor for unusual sign-in patterns, such as high-volume attempts from single sources, and automate responses to block them.
• Review access logs for any unauthorized entries during the period of the ConnectWise breach if you are a customer.
• Adopt a zero-trust model where service-to-service connections require continuous verification, not just initial credentials.
• Stay informed about open-source AI security tools like ARTEX to understand how they might be repurposed for attacks against your systems.