CybersecurityIn the news

What the FBI ARTEX Alert Asks Federal Security Teams to Check

Yahia Aboushadi
Director of Brand and Marketing, Pyramid Systems
Reading time:
6 min.

The FBI's November 13, 2026 FLASH on ARTEX, numbered FLASH-20261113-001, says some intrusions moved "from initial targeting to privileged access in under three hours." For federal IT teams, the more useful parts of the FBI ARTEX alert are where it says the tool succeeds and the list of logs it wants to see with a report.

What does the FBI ARTEX alert say?

The FBI says threat actors have weaponized ARTEX, an open-source penetration testing framework that uses agents driven by large language models (LLMs) to automate "multiple phases of cyber intrusions at machine speed, from reconnaissance through data exfiltration." Since at least August 2026 it has been used against organizations in many sectors, government among them.

ARTEX has no model of its own. It connects to outside LLM providers, commercial and open source, and the FBI says it lets actors "with minimal technical expertise" run complex campaigns against many targets at once.

The developer made the GitHub repository unavailable on October 8, 2026 after reports of abuse, but the FBI says forks, archived copies and a commercial version, ARTEX PRO, keep it in circulation. As of early November, researchers had found more than 1,600 publicly exposed ARTEX consoles.

The FLASH highlights two campaigns: attacks on South Korean financial institutions that exfiltrated personal information on about 67,000 people, and a supply chain compromise of ConnectWise Cloud Backup customers that reached more than 2,250 downstream tenants.

Where ARTEX gets in

ARTEX does best on systems that get less security attention than flagship applications. The FBI's list is specific: "partner and supplier portals, mobile backends, employee applications, sales-support systems, legacy services, third-party integrations, and applications created across different business units."

In a federal environment, those are often the systems that sit at the edge of the inventory: a grants or vendor portal built years ago by one office, an integration a contractor maintains. They may have an authorization on paper and still get little day-to-day scrutiny.

Much of the initial access the FBI observed is old weakness found faster. Its list includes unsigned SAML assertions, XML external entity (XXE) flaws, server-side request forgery (SSRF), missing authentication on web APIs, hardcoded administrator passwords in JavaScript bundles, insecure direct object references, weak administrative credentials, and stolen VPN configurations and leaked credentials. The FBI also says the agents can "map trust relationships between organizations" and pivot from a weaker supplier, CI/CD pipeline or software provider into the larger network.

What does the FBI tell defenders to do?

Mostly existing practice. The left column paraphrases the FBI; the right column is the evidence we'd want on file.

FBI recommendation Evidence we'd want on file
Inventory all internet-facing applications, including peripheral ones An inventory that names the portals, integrations and business-unit apps, with an owner for each
Restrict cloud administration, Kudu/SCM publishing endpoints, hypervisors and backup platforms to trusted networks; disable basic publishing authentication where not required Network rules showing those interfaces aren't reachable from the internet, and the publishing authentication setting for each app
Require architecture review before workloads or management interfaces are exposed to the internet A review record before each workload or management interface is exposed
Move enterprise applications behind Zero Trust access where feasible Which applications still face the internet directly, and why
Apply least privilege to multi-tenant app permissions and consent grants; monitor credential additions Current consent grants and an alert on new application credentials
Rate-limit portals and alert on sequential lookups; enforce server-side authorization Rate limits per endpoint and the alert that fires on high record counts per session
Scan code, pipelines, images and infrastructure-as-code for embedded secrets; rotate exposed ones Scan results and a rotation log

For the ConnectWise campaign, the FBI says to review every Microsoft 365 tenant where the affected ConnectWise applications are or were installed, "search by application ID rather than display name," and review activity from at least October 1, 2026. The FLASH doesn't list the application IDs themselves. It also says to keep those connections disabled until ConnectWise provides restoration instructions and not to delete enterprise applications as a first response, so evidence survives.

Why do AI pilots complicate detection?

Because one of the FBI's behavioral indicators is outbound API calls to LLM providers, which the FLASH says to "investigate unexpected calls alongside other suspicious activity." An agency running sanctioned AI pilots may see some of that traffic every day.

The word that matters is "unexpected." That check only works if someone keeps a current list of the exact hostnames each approved system calls and the gateway it uses. The FBI's indicators mix four provider API hosts with a reseller domain, xcai[.]pro, and its report list asks about "self-hosted LLM gateway or reseller software" and "token consumption patterns or API billing anomalies," so the AI program office and the security operations center should work from one list.

How to spot it, and the logs to have ready

The FBI says ARTEX generates "forensically noisy traffic." Its list of attack characteristics includes compressed bursts of activity, high volumes of failed attempts with iterative, incremental changes, and sub-agents overwriting each other or sending duplicate requests. The IP addresses in the FLASH are "historically observed infrastructure" that may since have been reassigned, so the FBI says to check them against current telemetry before blocking.

The FLASH's list of what to include in a report doubles as a logging test. It asks for service principal sign-in logs, Microsoft Graph and Entra audit records, sequential lookups and per-session record counts on partner and employee portals, outbound connections to LLM endpoints, cloud IAM activity, cloud metadata service queries, and evidence of web shells or new OAuth application credentials.

It also asks for "time from initial suspicious activity to detection" and "time from detection to containment." When the FBI says some intrusions reached privileged access in under three hours, those two numbers are the honest measure of readiness. Agency logging plans under OMB M-26-14 are due November 18, 2026, and our M-26-14 post covers what agencies are already required to log.

Where to start

Pull the inventory of internet-facing systems and add the peripheral ones the FBI names, each with an owner. Check that administrative and publishing interfaces aren't reachable from the internet. Build or update the list of systems approved to call LLM APIs, and hand it to whoever watches egress. Run a secrets scan across repositories and pipelines for the systems at the edge.

The FLASH sets no deadline of its own, and it says organizations have "no obligation to provide information" in response. Federal civilian agencies still report incidents to CISA under its notification guidelines, which our post on the CISA incident reporting form explains. Contractors running agency systems should ask the contracting officer or agency security office how the agency wants findings reported.

The Bottom Line

Most of the openings the FBI ARTEX alert lists are long-standing weaknesses on systems at the edge of an inventory, found and chained faster. Start with that inventory and the logs the FBI asks for, and treat anything you can't produce as the first finding. Our cybersecurity teams support agencies with the FISMA and NIST work behind both.

FAQ

What is ARTEX?

The FBI describes ARTEX as an autonomous, multi-agent penetration testing platform written in Go that turns goals written in plain language into commands executed by agents. It has no model of its own and connects to outside large language model providers, both commercial and open source. The developer said it was meant for authorized testing, but the FBI says threat actors have weaponized it.

Is ARTEX still available after the GitHub repository was removed?

Yes, according to the FBI. The FLASH says the framework continues to spread through community forks, localized variations, archived copies and a commercial version, ARTEX PRO, and remains available through mirrors on code-sharing platforms.

Does the FBI ARTEX alert set a deadline for federal agencies?

No. The FLASH is an FBI threat advisory, not a CISA directive. It names no CVEs and sets no deadline of its own, though any CVE already in CISA's Known Exploited Vulnerabilities catalog keeps its own due date. How an agency or contractor responds is set by its security office and, for contractors, the contract.

Where do you report suspected ARTEX activity?

The FLASH asks recipients to contact their local FBI field office or the Internet Crime Complaint Center (IC3) at ic3.gov, and lists the information that helps, from console screenshots to sign-in logs and LLM API usage. Federal civilian agencies also report incidents to CISA within one hour under CISA's Federal Incident Notification Guidelines; the FLASH doesn't address that process.

Sources

  1. FBI, FLASH-20261113-001, Threat Overview of ARTEX Autonomous Penetration Testing Framework (November 13, 2026)
  2. CISA, Federal Incident Notification Guidelines
  3. The Record, FBI flash alert says Artex tool automating cyberattacks at 'machine speed' (November 16, 2026)

How we researched this. Every number, date and requirement here is checked against the sources listed above, primary documents first, and the post is fact-checked again before publication. When a rule, date or status changes, we update the post and log the change below. Read our editorial standards or send a correction.

Yahia Aboushadi
About the author. Yahia Aboushadi is Director of Brand and Marketing at Pyramid Systems, an employee-owned federal IT firm in Fairfax, Virginia, that has delivered for federal agencies since 1995. . Connect on LinkedIn.

Continue reading

Related Insights

More from Pyramid Systems on federal cybersecurity.

Several monitors showing green code above a desk in a dark room

Compliance

16 November 2026
What DoD Checks Before Granting a Continuous ATO (cATO)

Continuous ATO from the primary documents: DoD's three cATO competencies, what the evaluation criteria require, who approves, and civilian ongoing authorization.

Read more
Network equipment glowing green in a dark server room, the kind of cloud infrastructure the FedRAMP VDR/VER migration covers

Compliance

14 November 2026
FedRAMP VDR/VER Migration: The December Progress Report, Then Three 2027 Dates

FedRAMP's VDR/VER migration is due December 7, 2026. What NTC-0017 asks for in the progress report, and the 2027 dates that lead to remediation or revocation.

Read more
Several monitors showing code in a dim security workspace

Cybersecurity

13 November 2026
CISA's New Incident Reporting Form: What Changed on November 9, 2026

CISA's incident reporting form changed on November 9, 2026. What the new questions ask, who must report within one hour, and what to have ready before you file.

Read more
Package installation output in green text on a dark Linux terminal, the kind of update that fixes SSHole CVE-2026-123158

Cybersecurity

10 November 2026
Is Your Alpine sshd Exposed to SSHole, CVE-2026-123158? A Four-Condition Check

SSHole, CVE-2026-123158, lets attackers log in to some musl-based Linux hosts with no password. The four conditions, the November 12 KEV date and how to check.

Read more