ARTEX: FBI Warns of AI Attack Tool That Can Break In Within Hours
The FBI says attackers are using ARTEX, an open-source AI framework, to automate intrusions, with some reaching privileged access in under three hours. Here is how it works and what businesses should do.
Image: TeQBlogs
The FBI has issued a FLASH alert about ARTEX, an open-source AI tool that attackers are using to run cyber intrusions automatically. The alert, dated 13 November 2026, says ARTEX uses AI agents to plan and carry out attacks with little human input, and that some intrusions reached privileged access within three hours of the first probe. It has been used against organisations worldwide since at least August 2026, so Indian businesses and their IT partners should treat this as a prompt to check their exposed systems now.
Quick summary
- What: FBI FLASH-20261113-001 on the ARTEX autonomous penetration testing framework.
- What ARTEX is: an open-source, AI-driven multi-agent tool built for security testing, now misused by attackers.
- Since when: malicious use seen since at least August 2026.
- Speed: some intrusions reached privileged access within three hours.
- Targets: finance, government, technology, defence, semiconductors, energy, healthcare, manufacturing, telecom and legal, worldwide.
- Scale: researchers found more than 1,600 publicly exposed ARTEX consoles as of early November 2026.

What is ARTEX?
ARTEX is a penetration testing platform written in the Go language. A user types a goal in plain language, and a team of AI agents turns it into commands: scanning, finding a weakness, exploiting it and moving deeper into the network. It has no AI model of its own. It connects to outside large language model services, and the FBI says the instances run by attackers mostly use Chinese-origin models.
According to the FBI, the developer published ARTEX on GitHub in July 2026 and removed the repository on 8 October 2026 after reports of abuse. That did not stop it. Copies survive as forks, mirrors and archives, and there is a commercial version called ARTEX PRO.
Also read: AI Cyberattacks Go Autonomous: Google Finds an Agent Turning Patches Into Exploits
Why is the FBI warning about ARTEX?
The concern is speed and scale, not a new kind of attack. The techniques are familiar ones. What changes is that software now chains them together in hours, and that a low-skilled attacker can run several campaigns at once.
- Fast: some intrusions went from first targeting to privileged access in under three hours.
- Good at finding weak edges: partner portals, mobile app backends, legacy services and third-party integrations.
- Supply-chain aware: agents can map trust between organisations and move from a weaker supplier into a larger company.
- Noisy but quick: its traffic may trigger alerts, yet defenders get very little time to respond.
The FBI cites two cases. Attacks on South Korean financial institutions exposed the personal information of about 67,000 people. A compromise involving ConnectWise Cloud Backup customers gave attackers access to more than 2,250 downstream tenants.
How an ARTEX attack unfolds
| Stage | What the FBI observed |
|---|---|
| Reconnaissance | Vulnerability scanning, web directory fuzzing, and testing what stolen cloud credentials are allowed to do |
| Initial access | Exploiting public-facing apps (SQL injection, SSRF, authentication bypass, open APIs), stolen VPN or vendor credentials, passwords hardcoded in JavaScript |
| Foothold | Web shells, added SSH keys, new OAuth credentials, open-source remote access tools |
| Privilege escalation | Kerberoasting, DCSync, cloud roles with too much power, forged admin tokens |
| Lateral movement | Pass-the-Ticket, WinRM, Kerberos relay and SMB relay |
| Collection | Mass reads of databases, storage and mailboxes, with output sent to rotating web endpoints |
| Command and control | Residential proxies, Tor and tunnelling tools |
Warning signs to look for
- Unexpected outbound connections from servers to AI model APIs that your business does not use.
- High-frequency, automated requests that step through IDs or records one after another on login or search pages.
- Web shell files (
.aspxor.jsp) in application folders where they do not belong. - Large outbound transfers of structured personal or financial data to unfamiliar destinations.
- New credentials, owners or permissions on cloud applications that nobody on your team added.
The FLASH also lists IP addresses and a domain linked to the campaigns. The FBI advises checking these against your own current logs before blocking, because IP addresses get reassigned.
What businesses should do now

- List everything that faces the internet. Include forgotten portals, test sites and systems built by individual departments.
- Lock down admin interfaces. Cloud consoles, network management, hypervisors and backup platforms should be reachable only from trusted networks.
- Apply least privilege to cloud apps, service accounts and consent grants, and review them regularly.
- Rate-limit important APIs and portals, and alert on sequential lookups or unusually high record counts in one session.
- Hide error details. Do not show stack traces, SQL errors or internal paths to visitors who are not signed in.
- Hunt for secrets. Scan code, configuration files, CI/CD pipelines and containers for embedded passwords and keys, and rotate anything exposed.
- Watch identity logs for new credentials, role changes and unfamiliar sign-ins by applications.
- Preserve evidence, but contain first. Keep sign-in and audit logs, and do not delay urgent containment while waiting for exports.
Organisations that use ConnectWise Cloud Backup get specific guidance in the FLASH: review every Microsoft 365 tenant where the affected applications were authorised, look at activity from at least 1 October 2026, and keep the application connections disabled until ConnectWise issues formal restoration instructions.
What this means for Indian businesses
The FBI says targets are worldwide, and the weak points ARTEX looks for are common in Indian small and mid-sized companies: an old vendor portal, a mobile app backend nobody patches, or a shared admin password. Managed service providers are a particular risk, because one compromised provider can open the door to many clients. Ask your IT partner which of your systems are exposed to the internet and how quickly they would spot the signs listed above.
In India, cyber incidents should also be reported to CERT-In. If you need help with an exposure review, Suprams’ cybersecurity team works with businesses on this. Related reading: how to secure AI agents, the Cursor security breach, the Citrix NetScaler bulletin and CERT-In’s Crafty Falcon warning.
Frequently asked questions
What is ARTEX?
ARTEX is an open-source, AI-driven penetration testing framework that uses multiple AI agents to automate attack steps. It was built for security testing but is being misused by attackers.
Is ARTEX malware?
Not in the usual sense. It is a tool that runs on the attacker’s side and drives the attack. The damage comes from what it does to exposed systems, such as planting web shells and stealing data.
Who is being targeted?
The FBI lists financial services, government, technology, defence, semiconductor, energy, healthcare, manufacturing, telecommunications and legal organisations worldwide.
How fast are ARTEX attacks?
The FBI says some intrusions reached privileged access within three hours of initial targeting.
The bottom line
ARTEX does not use new tricks. It makes old ones fast and cheap. The defence is the same basic hygiene done properly: know what you expose to the internet, keep admin panels private, remove hardcoded secrets, limit permissions and watch your identity logs. Start with the internet-facing systems you have not looked at in a while.
