The FBI Warned About ARTEX. We Use It to Find Your Weaknesses First.

The FBI warned that attackers use the ARTEX AI framework to break in at machine speed. ITECS runs the same framework on its own models in its Dallas data center, under signed rules, to find your weaknesses before attackers do.

Back to Blog
9 min read
Headline “Find your weaknesses before attackers do” above a glowing AI core sealed in a shield-shaped glass vault in a data center, with scoped blue links to client systems and red AI streaks held outside.

On November 13, 2026, the FBI warned that a variety of malicious cyber actors are using an AI framework called ARTEX to automate cyberattacks. According to the Bureau, some intrusions went from first targeting to privileged access in under three hours. For business owners, boards and executive teams, the question is no longer whether attackers have machine-speed tools. It is whether your organization finds its own weaknesses first.

ITECS uses the same framework to do exactly that. We run ARTEX inside our own Dallas data center, connected only to AI models we host ourselves, under rules that you sign off before any test begins. Attackers use it to break in. We use it, safely and with your authorization, to show you where they would.

Key points

  • The FBI says ARTEX lets attackers with little expertise run sophisticated, multi-target intrusions at machine speed, and that it is especially good at finding overlooked systems.
  • Attackers are fast; most organizations are not. Verizon found the median time to fully fix a known-exploited critical vulnerability was 43 days.
  • ITECS runs ARTEX on private models in its Dallas data center. Nothing from a test is sent to a public AI service, and your report is delivered to you securely.
  • Every test runs under signed rules of engagement, and our testers reproduce each AI finding by hand before it reaches your report.

Under 3 hours

from first targeting to privileged access in some ARTEX intrusions

43 days

median time to fully fix a known-exploited critical vulnerability in 2025

31%

of breaches with a known entry point started with an exploited vulnerability, the most common way in

1 in 4

malicious breaches were AI-enabled, costing about $6 million on average

Sources: FBI FLASH-20261113-001 (November 13, 2026); Verizon 2026 Data Breach Investigations Report; IBM 2026 Cost of a Data Breach Report.

What the FBI warned about

The FBI’s FLASH advisory describes ARTEX as an open-source, AI-driven penetration testing platform. A user describes a goal in plain language, and the framework’s AI agents turn it into the technical steps of an intrusion, from reconnaissance through data theft, with little human involvement. ARTEX has no AI model of its own; it uses whatever AI model its operator connects. The FBI notes that attackers mostly pair it with low-cost models that show no observable guardrails when a request is framed as penetration testing.

The developer, who said the tool was intended for authorized security testing within defined scopes, released ARTEX in July 2026 and removed the original repository on October 8 after reports of abuse. Copies, forks and a commercial edition remain in circulation, and researchers found more than 1,600 publicly exposed ARTEX consoles in early November. The FBI links ARTEX to campaigns that stole personal information on about 67,000 people from South Korean financial institutions, and to a supply chain compromise that led to attempted sign-ins at more than 2,250 downstream customer environments.

Three findings in the advisory matter most to leadership:

  • A lower barrier to entry. Small groups with minimal expertise can now run complex intrusions against many targets at once.
  • A focus on what you overlooked. ARTEX excels at finding peripheral systems: partner and supplier portals, employee applications, mobile backends, legacy services and third-party integrations.
  • Less time to react. The attack moves fast enough that defenders have little time to respond, even when alerts fire.

The speed gap your board should see

AI has made attackers faster. Most organizations have not become faster at fixing what attackers exploit.

ATTACKER WITH ARTEX

Under 3 hours

From first targeting to privileged access in some intrusions the FBI cited.

TYPICAL ORGANIZATION

43 days

Median time to fully fix a critical vulnerability already known to be exploited.

Verizon’s 2026 Data Breach Investigations Report found that exploiting a vulnerability is now the most common way breaches begin, at 31% of breaches with a known entry point. Only 26% of critical, known-exploited vulnerabilities were fully fixed in 2025, down from 38% the year before, while organizations had 50% more of them to patch. Breaches involving a third party rose 60% to 48% of all breaches.

Mandiant’s M-Trends 2026 tells the same story. Exploits were the most common way in for the sixth year running, at 32% of intrusions, and attackers stayed inside for a median of 14 days before discovery. The median time between one criminal group gaining access and handing it to another collapsed from more than eight hours in 2022 to 22 seconds in 2025.

The cost is rising too. IBM’s 2026 Cost of a Data Breach Report found that one in four malicious breaches was AI-enabled, a 56% increase, and those breaches cost about $6 million on average against a global average of $4.99 million.

Same engine, different rules

The technology itself is neither good nor bad. What separates an attack from a security test is authorization, scope, control of the AI and what happens to the data. This is how ITECS runs the same framework the FBI warned about:

How attackers use ARTEX How ITECS uses it
Permission None Your signed rules of engagement, before any test begins
Targets Anything reachable Only approved targets, during approved testing windows; excluded systems stay off-limits
The software Forks, mirrors and a commercial edition of unknown provenance Our own audited build of the original release, with no backdoors and no way to reach an unauthorized endpoint
The AI model Mostly low-cost outside models with no observable guardrails Models we host ourselves, on ITECS-owned GPUs in our Dallas data center
Risky actions Whatever achieves the goal Hardwired to non-destructive testing methods; any higher-risk action still needs your explicit approval
Your data Stolen and sent to attacker-controlled destinations Stays in the ITECS data center; never sent to a public AI service or used to train anyone else’s model
Human oversight Minimal by design Our testers reproduce every finding by hand before it is reported
The outcome A breach A secure report, a remediation plan and a retest

The software matters as much as the model. The original ARTEX repository is gone, and the copies attackers circulate come from forks, mirrors and a commercial edition. ITECS works from a copy of the original release that we audited and modified ourselves, so it has no backdoors and no way of reaching out to an unauthorized endpoint. We extended it to test for common and known vulnerabilities (published CVEs), and hardwired safeguards that limit it to non-destructive methods in your environment.

Running the models ourselves is what makes that control possible. Because the AI never leaves our infrastructure, we decide how it is configured, what it may do and what it may not, for every engagement. It also means our testing makes no calls to the public AI services that the FBI lists among the warning signs of ARTEX activity.

What governed AI testing covers

ITECS applies the same AI-assisted approach across the places attackers look first, including the overlooked systems the FBI highlighted:

Software and applications

Business-logic and access-control flaws in the applications your teams and customers rely on.

APIs

The interfaces behind your apps, portals and integrations, where one weak check can expose many records.

Websites and portals

Public sites plus the partner, supplier and employee portals that often receive less attention.

Internal networks

Weak segmentation and the paths an attacker would use to move from one foothold to your most sensitive systems.

External edge

Internet-facing services, remote access and exposed management interfaces at the edge of your network.

Known vulnerabilities

Broad vulnerability assessment that finds and prioritizes what to fix first.

How a governed AI test works

Every engagement follows the same core stages, with people in control at each one:

1Scope and authorize

You approve the targets, exclusions and testing windows, which we coordinate with your IT and security team, and no high-risk action runs without your explicit approval. Nothing runs before the rules of engagement are signed.

2Map the attack surface at machine speed

ARTEX and our private models take on the discovery and triage that used to consume days, inside the approved scope only.

3Prove what matters

Our testers reproduce each finding by hand and focus their time on business logic and chained attacks, so you never chase a hallucinated vulnerability.

4Report securely

Critical findings are reported as soon as a tester confirms them. Your executive summary, technical report and remediation workbook are handed over securely on the date set at scoping.

5Fix and retest

A remediation workshop helps your team close the gaps, and retesting is included within the window of your package.

Your data never leaves our data center

Stays inside ITECS

  • The AI models and every prompt
  • Credentials and target data
  • Draft findings and test evidence

Delivered to you, securely

  • Executive summary for leadership
  • Technical report and remediation workbook
  • Retest results after you fix

Questions your board should ask

Whether or not you work with ITECS, these questions show whether your organization is ready for machine-speed attackers:

Board and executive checklist

  • Do we have a current inventory of every internet-facing system, including partner portals, employee apps and legacy services?
  • When were our applications, APIs, networks and external edge last tested, and how?
  • How long do we take to fix vulnerabilities that attackers are already exploiting?
  • If a security provider uses AI, where does the model run, and where does our data go?
  • Who authorizes each test, and how are its scope and high-risk actions controlled?
  • Who proves each finding before it reaches our report, and is a retest included?

The bottom line

ARTEX shows what attackers can now do: test many doors at machine speed and walk through the first one left open. The answer is not to avoid the technology. It is to use it first, under your authority, on private infrastructure, with people proving every result. That is what ITECS does, so you find your weaknesses before attackers do.

Find your weaknesses before attackers do

Governed, AI-powered penetration testing from ITECS, run on private models in our Dallas data center.

Schedule an AI-powered penetration test

Testing a specific surface? See web application, API and vulnerability assessment services.

Sources and methodology

Methodology: Threat details and statistics come from the primary sources above. The FBI does not endorse any company, product or service. Descriptions of ITECS models, testing controls and deliverables reflect current ITECS operations. Prepared by the ITECS team with AI-assisted research on November 17, 2026.

continue reading

More ITECS blog articles

Browse all articles

About ITECS Team

The ITECS team consists of experienced IT professionals dedicated to delivering enterprise-grade technology solutions and insights to businesses in Dallas and beyond.

View full profile and articles

Share This Article

Continue Reading

Explore more insights and technology trends from ITECS

View All Articles