FBI Warns Hackers Are Using Open-Source AI Framework ARTEX to Breach Banks and Technology Firms Within Hours
Key Takeaways
- •The FBI warned on November 13 that attackers are using ARTEX, an open-source AI framework designed for penetration testing, to automate cyberattacks against banks, technology companies, and suppliers.
- •Some intrusions powered by the framework advanced from target selection to privileged access in less than three hours, sharply reducing the time security teams had to respond.
- •Attacks on South Korean financial institutions tied to the framework exposed the personal information of approximately 67,000 people.
- •Microsoft connected ARTEX to a ConnectWise Cloud Backup supply-chain compromise involving Storm-0940, a China-aligned group that attempted sign-ins across more than 2,250 customer environments using compromised application identities.
- •Microsoft advised affected customers to examine application sign-ins, investigate unexpected permission changes, and preserve evidence, while noting that successful authentication alone does not confirm that customer data was accessed or stolen.

The FBI warned on November 13 that hackers are leveraging ARTEX, an open-source artificial intelligence framework, to automate attacks against banks, technology companies, and suppliers. According to the agency, some intrusions advanced from target selection to privileged access (high-level permissions over systems and data) in less than three hours — a pace that sharply reduced the time security teams had to respond.
The advisory stated that multiple attackers have used the framework since at least August 2026 across financial services and other industries. The FBI cited attacks against South Korean financial institutions that exposed the personal information of approximately 67,000 people, underscoring the potential consequences for lenders and their customers.
AI Built for Testing, Turned Against Corporations
ARTEX was originally designed for authorized penetration testing, which helps organizations identify weaknesses before criminals exploit them. Operators describe objectives in ordinary language, and AI agents translate those instructions into technical actions. Rather than relying on its own model, the framework connects to external AI models. Its open-source code is publicly available, and because it draws on those outside models, operators do not need to build or maintain an AI system of their own to use it.
The FBI noted that the same capabilities can be repurposed by attackers to investigate networks, exploit vulnerabilities, and pursue multiple targets simultaneously. The agency also highlighted frequently overlooked systems, including supplier portals, employee applications, and older services. For businesses, such connections can give criminals a route into more valuable corporate networks through a weaker partner.
ConnectWise Supply-Chain Compromise Expands Exposure
Microsoft's investigation linked ARTEX to a ConnectWise Cloud Backup supply-chain compromise involving Storm-0940, a group aligned with the Chinese state. The attackers used compromised application identities — credentials assigned to software rather than to individual employees — to attempt sign-ins across more than 2,250 customer environments, taking advantage of existing permissions that allowed the applications to connect without an interactive user login.
Microsoft cautioned that successful authentication alone does not establish that attackers accessed or stole customer data. Nevertheless, investigators observed follow-on activity in multiple environments, including credential changes and attempts to expand privileges — actions that can help intruders retain access after defenders address the original entry point.
Shrinking Response Windows
Although automated attacks can generate conspicuous traffic, their speed leaves defenders little time to coordinate containment. Microsoft urged affected customers to examine application sign-ins, investigate unexpected permission changes, and preserve evidence. With attackers reported to have used the framework since at least August 2026, that review of sign-ins and permissions reaches back well before the public warning itself. The guidance underscores the need to scrutinize trusted software connections alongside employees' accounts when investigating a potential breach.
Source: CryptoMeter io