AW Dev Rethought Flash: FBI Warns ARTEX AI Tool Is Breaching Networks in Under 3 Hours
The FBI has issued a public warning about ARTEX, an open-source AI tool built for security testing that attackers are now using to break into organisations worldwide. The alert, FLASH-20261113-001, was released on Friday, November 13, through the Internet Crime Complaint Center.
The Bureau says ARTEX has been used in attacks since at least August 2026. In some cases, intruders went from first contact to privileged access in under three hours.
What ARTEX Is
ARTEX is an autonomous penetration testing framework written in Go. An operator gives it a goal in plain language. A planner turns that goal into tasks, and a team of AI agents carries them out with little human input.
Those agents handle most of an intrusion themselves: scanning targets, finding weaknesses, exploiting them, escalating privileges and collecting data.
ARTEX has no AI model of its own. It connects to external ones, including Anthropic’s Claude, OpenAI’s GPT, DeepSeek, GLM, Grok and Kimi. The FBI says attacker-run instances most often use Chinese-developed models, led by the DeepSeek Flash family.
How a Testing Tool Became an Attack Tool
A developer using the handle “Autumn-27” published ARTEX on GitHub in July 2026. The developer said it was meant for authorised testing.
Reports of abuse followed, and the repository was taken down on October 8. That did not remove the tool. The FBI says it still circulates through forks, archived copies, localised versions and a commercial edition called ARTEX PRO.
Researchers counted more than 1,600 ARTEX control consoles exposed on the internet in early November. The FBI believes the real number is much higher.
Taking the repository down did not take the tool away. The copies were already out.
Who Has Been Hit
The alert does not blame a single group. It describes a variety of actors using the same tool across finance, government, technology, defence, semiconductors, energy, healthcare, manufacturing, telecoms and legal firms.
Three campaigns are tied to it in the alert:
- South Korean banks. A financially motivated campaign stole personal information on about 67,000 people. CrowdStrike and Zscaler assess, with moderate confidence, that the operator was a Chinese speaker.
- ConnectWise Cloud Backup. Attackers abused a trusted ConnectWise application to go after more than 2,250 downstream Microsoft 365 tenants. Microsoft attributes it to Storm-0940, which it describes as China-state aligned. Huntress reported at least 162 compromised tenants.
- Storm-3358. Microsoft tracked ARTEX infrastructure for this group that was still active on November 6.
BleepingComputer reports two more cases. CrowdStrike says a group it calls Vapor Panda used ARTEX against more than 200 organisations in 14 countries, compromising at least nine. South Korean police arrested seven suspects over attacks on more than 90 organisations, with about 8.2 million records stolen.
The Attacks Are Not New. The Speed Is.
The alert cites no zero-days and no specific CVEs. ARTEX works with ordinary weaknesses: hard-coded passwords, unrotated keys, exposed credentials, SQL injection and weak authentication.
It goes after the systems that get the least attention. Partner portals, employee apps, legacy services and third-party integrations are the common entry points. It also maps trust between companies, so a weaker vendor becomes a route into its customers.
Akamai’s summary, quoted in coverage of the alert: AI “changes the speed of attack, not the fundamentals.”
The FBI lists the consequences. Attackers with little skill can now run complex operations against many targets at once. Defenders may see the alerts, but have far less time to act on them.
One detail favours defenders. In a case described in the alert, working endpoint protection blocked the agents’ attempts to move between machines.
What the FBI Recommends
- Inventory every internet-facing application, including partner and business-unit systems.
- Take cloud administration and backup interfaces off the public internet.
- Apply least privilege to multi-tenant app permissions and consent grants.
- Rate-limit sensitive APIs and alert on rapid sequential lookups.
- Stop returning stack traces and SQL errors to unauthenticated users.
- Scan code, pipelines and container images for embedded secrets, and rotate any that are exposed.
- Watch for servers making unexpected calls to LLM APIs. That is now a useful signal.
- ConnectWise Cloud Backup customers should review Microsoft 365 tenant activity from at least October 1.
The FBI also warns that the IP addresses in its indicator list may have been reassigned. Check them against your own logs before blocking.
Suspicious activity can be reported to a local FBI field office or to IC3.
Enjoyed this post?
Stay in the loop
New posts + weekly digest, straight to your inbox.
Create a free account
- Save posts to your vault
- Like posts & build history
- New-post alerts
No comments yet. Be the first to comment!