The FBI has issued a FLASH advisory warning that threat actors are reportedly using an artificial intelligence-enabled penetration-testing framework known as ARTEX to speed up attacks against organizations in multiple sectors and countries.
According to the bureau, the tool has been misused since at least August 2026 in activity affecting financial services, government, technology, defense and healthcare targets. The FBI said the platform can automate portions of reconnaissance, vulnerability discovery, exploitation, privilege escalation and data collection, potentially allowing operators to run campaigns against several targets at once.
Automation and LLM integrations
ARTEX is written in Go and uses specialized agents directed by user-provided objectives. It does not rely on a proprietary model; instead, it can connect to external large language models, the FBI said. The agency warned that this structure can lower the technical barrier for conducting complex intrusion sequences and compress the time between initial scanning and access to privileged systems.
The framework was reportedly published on GitHub in July by a developer using the name Autumn-27, then removed in October amid reports of criminal abuse. The FBI said copies remain accessible through forks, archives, localized variants and a commercial product described as ARTEX PRO. Researchers had identified more than 1,600 exposed management consoles by early November, although the total number of installations is uncertain.
Campaigns cited in advisory
The advisory references several reported incidents linked to the framework. These include the October compromise of ConnectWise Cloud Backup application identities, which reportedly enabled attempts to access downstream Microsoft 365 tenants. Microsoft previously attributed that activity to a China-aligned group it tracks as Storm-0940, while Huntress reported finding compromised customer tenants during its investigation.
The FBI also cited reported attacks on South Korean financial institutions and a campaign attributed by CrowdStrike to the suspected China-linked group Vapor Panda. In those cases, ARTEX was allegedly used alongside other tools and services to automate discovery, credential attacks and follow-on access attempts.
Recommended defensive actions
Organizations should prioritize internet-facing applications, cloud administration interfaces, third-party integrations and legacy services, the FBI said. Recommended measures include reviewing application permissions and authentication records, limiting administrative access, applying rate controls to sensitive APIs, and watching for bursts of automated requests, unexpected privilege changes, unusual cloud activity and large outbound data transfers.
The advisory contains indicators of compromise, but the FBI cautioned that infrastructure-based indicators can change quickly. Organizations that identify potentially related activity are encouraged to preserve relevant logs and report findings to the FBI or the Internet Crime Complaint Center.
