FBI Reports ARTEX AI Tool Enabled Hackers to Secure Privileged Access in Less Than 3 Hours

The FBI has issued a warning regarding a sophisticated new cyberattack framework, known as ARTEX, which automates critical stages of cyberattacks, including target selection and data theft.

As outlined in a Flash Advisory on November 13, the Federal Bureau of Investigation (FBI) reported that cybercriminals have been exploiting this open-source framework since at least August 2026 to launch attacks against a variety of organizations globally. Sectors under threat include financial services, healthcare, government, technology, defense, energy, and legal services.

One notable incident involved a compromise in the supply chain of ConnectWise Cloud Backup, allowing intruders to infiltrate more than 2,250 downstream clients.

Developed in Go, ARTEX connects to leading large language models (LLMs) such as DeepSeek, Claude, and GPT. Its innovative approach features specialized AI agents that take on distinct tasks once an operator designates an objective. These agents can autonomously identify potential targets, assess system vulnerabilities, exploit weaknesses, and extract data with minimal human intervention.

The FBI noted that certain attacks can progress from the initial targeting phase to gaining privileged access in under three hours. Attackers favor DeepSeek Flash models due to their cost-effectiveness and the absence of visible safety protocols for penetration testing, making them particularly appealing for malicious use.

The problem extends beyond rapid attacks

Some intrusions facilitated by ARTEX have achieved privileged access in less than three hours, providing security teams with limited time to respond before attackers deepen their infiltration within the network. The ConnectWise breach illustrates the ripple effect that such vulnerabilities can have across multiple organizations.

An investigation by Microsoft connected ARTEX activity to Storm-0940, a threat group purportedly linked to China. After successfully breaching ConnectWise applications, the attackers leveraged their access to target customer environments, generating over 151,000 failed sign-in attempts in a related operation that peaked at 2,900 attempts per minute.

The FBI has also underscored attacks aimed at South Korean financial institutions, revealing that personal data of approximately 67,000 individuals was compromised. Affected banks included Shinhan, KB Kookmin, and Hana, among others.

These instances underscore a significant shift in the manner attackers employ AI. Previous reports have detailed how ARTEX facilitated a campaign against PaperCut, wherein hundreds of AI agents assisted in exploiting two zero-day vulnerabilities. By consolidating multiple attack vectors into a singular framework, ARTEX reduces the complexity required to transition between various stages of an attack.

Google’s threat intelligence team has also identified the malicious use of alternative agentic frameworks, such as CAIRN and PentAGI, suggesting these operations, while potentially rudimentary and noisy, can still achieve success due to their rapid execution and scale.

Dismantling the original tool did not eliminate its use

The developer of ARTEX, Autumn-27, removed the original repository from GitHub on October 8, 2026, in response to reports of misuse. However, numerous copies, community forks, and a commercial version called ARTEX PRO remain accessible, with over 1,600 publicly exposed ARTEX consoles identified as of early November.

The FBI is actively seeking authentication records, API activity logs, and other evidence associated with ARTEX-related cyber incidents.

Concerns have also been raised regarding how organizations manage exposed systems, credentials, and access shared with suppliers. Morey J. Haber, Chief Security Advisor at BeyondTrust, noted that ARTEX diminishes the barriers of time and expertise required to exploit discovered weaknesses.

Haber emphasized that the real issue lies not in new exploits or zero-day vulnerabilities proliferated by AI, but rather in how quickly and effectively attackers can weaponize vulnerabilities. Successful breaches can occur in less obvious areas, such as partner portals and outdated services. Previously, chaining these dependencies required significant effort; ARTEX simplifies this process.

Furthermore, Haber advised organizations to pay closer attention to the identities and credentials utilized by applications and automated services, rather than focusing solely on employee accounts. “ARTEX isn’t merely breaching defenses; it’s exploiting forgotten identities and unattended service accounts. The interconnectedness between organizations means that your risk is only as secure as your weakest supplier,” he stated, highlighting the expanded threat landscape organizations now face.

Source