🔍 Read the full analysis: FBI On ARTEX AI: What The “Under Three Hours” Phrase Does And Doesn’t Mean on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
An FBI warning dated November 13, 2026, says threat actors have used ARTEX, an AI-driven penetration-testing framework, in intrusions since at least August. Its “under three hours” figure comes from multiple Storm-0940 incidents cited by Microsoft; it is not a typical-case estimate, and a separate Barracuda incident lasted about 44 hours.
The FBI warned on November 13 that threat actors have used the open-source ARTEX framework to automate stages of cyber intrusions, with some cases reaching privileged access in under three hours. That timing is drawn from multiple incidents involving one group, not an average for ARTEX attacks; a separate Barracuda case described in the source material lasted about 44 hours.
The FBI’s 13-page FLASH advisory, FBI FLASH-20261113-001, describes ARTEX as an autonomous, multi-agent penetration-testing platform written in Go. Operators can set an objective in plain language, after which AI agents carry out tasks with limited further human input, according to the bureau. The advisory says ARTEX can support multiple phases, from reconnaissance through data exfiltration. The warning is marked TLP:CLEAR, meaning it can be shared without restriction.
ARTEX does not include its own AI model. The FBI lists integrations with external models including Claude, GPT, DeepSeek, GLM, Grok and Kimi. The bureau says Chinese models are most prevalent in attacker-controlled installations, with DeepSeek Flash models frequently used. These are descriptions in the advisory, not evidence that the model providers themselves took part in the intrusions.
The phrase “under three hours” refers to multiple instances in Microsoft’s October 21 account of Storm-0940, which Microsoft describes as China-state-aligned. Microsoft said the group moved from initial targeting to privileged cloud access in less than three hours in those instances, while also reporting that most targeting attempts were unsuccessful. The FBI’s wording does not establish that all ARTEX attacks, or even most of them, follow that timeline.
Why the Attack Clock Matters
The reported speed matters because security teams may have little time to investigate and contain a successful intrusion once an attacker begins working. The FBI describes a compressed detection window: ARTEX activity may produce noisy traffic and trigger alerts, but automation can let an attacker move quickly before an organization coordinates a response. The “under three hours” figure is a warning about what can happen, not a forecast of how long every incident will take.
The Barracuda incident illustrates why speed is only part of the picture. According to the source account, an intruder used a forgotten cloud key on November 1, 2026, and spent roughly 44 hours mapping access to about 10 cloud accounts. Barracuda engineers noticed the activity on November 4, after it had stopped. Barracuda said the pattern matched ARTEX. The case points to exposure from credentials and cloud access controls, even when an intrusion does not match the fastest reported timeline.
The FBI says ARTEX has been used against organizations across sectors, including finance, government, technology, defense, energy and healthcare. It highlights attacks on South Korean financial institutions that exposed personal information belonging to about 67,000 people, and a ConnectWise Cloud Backup supply-chain compromise that led to attempted sign-ins to more than 2,250 downstream tenants. Those figures describe specific incidents in the advisory, not the total impact of ARTEX activity.
Products Worth Considering

Samsung 32 Odyssey G5 QHD Gaming Monitor

Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite

Samsung 27-Inch Odyssey G5 QHD Gaming Monitor (G50F)
As an affiliate, we earn on qualifying purchases.
How ARTEX Reached Attackers
According to the FBI, a developer using the name Autumn-27 published ARTEX on GitHub in July 2026. The project reportedly accumulated an estimated 1,500 to 2,000 stars in about 10 weeks. Its developer said it was intended for authorized testing and removed the repository on October 8 after reports of abuse. The FBI says copies continued to circulate through forks, archived versions, localized variants and a commercial version called ARTEX PRO.
The timing claims in the source material come from different observations and should not be combined into one benchmark. Microsoft described multiple Storm-0940 cases under three hours from targeting to privileged access. Google’s threat intelligence team, citing Wiz research on a campaign overlapping with Storm-0940, reported a typical time of around four hours from the first compromised account to deep system access, with one case taking as little as 15 minutes. The Barracuda account, by contrast, describes roughly 44 hours of activity. These accounts use different starting points and milestones.
Amazon, which monitored two ARTEX campaigns through decoy sensors, said none of the techniques it observed were new. That observation suggests the concern is not necessarily a novel vulnerability: automation can help attackers apply established methods quickly and at scale. The FBI’s warning focuses on both the tool and the weaknesses it can exploit, including supplier portals, legacy services, mobile backends, employee applications and third-party integrations that may receive less security attention than flagship systems.
“Some intrusions went from initial targeting to privileged access in under three hours.”
— FBI, FLASH-20261113-001
What the Timing Cannot Establish
The available accounts do not show how often ARTEX attacks succeed, how often they reach privileged access, or what the typical duration is across all affected organizations. Microsoft’s “multiple instances” concerns one group’s activity, and the source says most of its targeting was unsuccessful. The reported four-hour and 15-minute figures from research cited by Google refer to different milestones than the FBI’s targeting-to-privileged-access measure.
The Barracuda account in the source material says the intruder used a key that should have been retired, but does not provide a complete public incident timeline or independently establish every technical detail. The FBI advisory is described as a bureau-issued warning; the source reports no joint CISA advisory. The full scope of victims, losses and ongoing investigations is not established in the supplied material.
Defenses and Follow-up Reporting
The FBI advisory directs defenders to its technical guidance and indicators, which the source account does not reproduce. Organizations can review internet-facing services, retire long-lived or forgotten credentials, restrict what each credential can access, rate-limit sensitive APIs and retain logs that can support an investigation. These steps address weaknesses identified in the reporting; they do not guarantee prevention of an intrusion.
Further clarity will depend on disclosures from affected organizations and any additional technical reporting from the FBI, Microsoft, Google, Amazon or Barracuda. Readers should watch for updated victim counts, confirmed incident timelines and evidence about how often attackers achieved privileged access. Until those details are available, “under three hours” is a documented possibility, not a general estimate of ARTEX incident speed.
Key Questions
What does the FBI mean by “under three hours”?
The phrase refers to multiple Storm-0940 incidents cited by Microsoft in which attackers moved from initial targeting to privileged cloud access in less than three hours. It is not a typical duration for every ARTEX incident.Does the figure mean ARTEX attacks usually succeed?
No. Microsoft said most of the targeting it observed was unsuccessful. The cited fast cases show what happened in some incidents, not an overall success rate.How long did the Barracuda incident take?
The source account says the activity lasted about 44 hours and involved mapping access to roughly 10 cloud accounts. Barracuda engineers noticed it after the activity had stopped.Is ARTEX itself an AI model?
No. The FBI describes ARTEX as an AI-driven framework that connects to external models. It does not have a model of its own.What should organizations review?
The source account points to reviewing exposed systems and third-party services, retiring unused credentials, limiting credential permissions, rate-limiting sensitive APIs and maintaining logs. The FBI advisory contains technical guidance and indicators not reproduced here.Source: ThorstenMeyerAI.com
