The Federal Bureau of Investigation has issued a formal advisory regarding the widespread exploitation of ARTEX, an autonomous penetration testing framework powered by artificial intelligence. Since August 2026, the tool has been utilized by various threat actors to automate cyberattacks across multiple critical sectors, including government, defense, finance, healthcare, and technology. The FBI warns that the framework significantly accelerates the attack lifecycle, allowing intruders to move from initial reconnaissance to privileged system access in under three hours.
ARTEX is a Go-based, open-source framework that functions by integrating with external large language models (LLMs) such as OpenAI GPT, Anthropic Claude, DeepSeek, GLM, Grok, and Kimi. Rather than relying on a native AI model, the framework translates user-defined objectives into actionable commands executed by specialized AI agents. These agents are capable of performing vulnerability discovery, privilege escalation, and data exfiltration with minimal human intervention. This automation enables even operators with limited technical expertise to conduct sophisticated, multi-target operations simultaneously.
The framework was originally published on GitHub in July 2026 by a developer identified as Autumn-27. Following reports of malicious abuse, the repository was removed in October 2026. Despite this, the FBI confirms that the tool remains active through various community forks, localized versions, and a commercial iteration known as ARTEX PRO. As of early November, security researchers had identified over 1,600 publicly exposed ARTEX consoles, though the total number of active deployments is suspected to be much higher.
The FBI advisory specifically links ARTEX to the October 2026 supply-chain attack against ConnectWise Cloud Backup, formerly known as SkyKick Cloud Backup. In this incident, attackers compromised application identities to gain unauthorized access to Microsoft 365 customer environments. Microsoft attributed the campaign to a China-state-aligned actor known as Storm-0940, which utilized ARTEX to conduct reconnaissance and password spraying against more than 2,250 downstream tenants. Investigations by cybersecurity firms confirmed that at least 162 tenants were compromised during this operation.
Additional campaigns identified by the FBI and private security researchers highlight the framework's global reach. In South Korea, a Chinese-speaking threat group combined ARTEX with Claude Code agents to target major financial institutions, including Shinhan Bank, KB Kookmin Bank, and Hana Bank. This campaign resulted in the exposure of personal information for approximately 67,000 individuals. Furthermore, South Korean authorities arrested seven suspects linked to a broader operation that compromised over 90 organizations and resulted in the theft of 8.2 million data records.
CrowdStrike has also documented the use of ARTEX by a Chinese cyberespionage group identified as Vapor Panda. This group reportedly targeted more than 200 organizations across 14 countries, successfully compromising at least nine of them. In these instances, the AI agents were observed performing automated reconnaissance, deploying web shells, and dynamically adjusting attack paths when initial exploitation attempts were unsuccessful.
To assist in defense, researchers have released an open-source detection signature pack designed to identify ARTEX-related activity. This includes rules for monitoring suspicious network requests, identity abuse, and automated attack patterns. The FBI recommends that organizations prioritize the security of internet-facing applications, restrict access to cloud administration interfaces, and implement rate limiting on sensitive APIs. Defenders are also advised to monitor for unusual bursts of automated traffic and unexpected changes to credentials or privilege levels.
While the FBI has provided a list of indicators of compromise, including IP addresses and command-and-control infrastructure, the agency cautions that these indicators may change rapidly. Organizations are encouraged to report any suspicious activity associated with the framework to the FBI’s Internet Crime Complaint Center. The bureau emphasizes that while the speed of ARTEX-powered attacks can overwhelm traditional security response times, proactive monitoring of application permissions and authentication logs remains a critical defense against these automated threats.





