First posted on Instagram
The Federal Bureau of Investigation has warned American businesses that cybercriminals are using artificial intelligence to break into computer networks in under three hours, reaching privileged access — the highest level of control inside a system — before the targeted company even realises it is under attack. The warning came in a FLASH alert issued on Friday, November 13, 2026, numbered FLASH-20261113-001, and it describes an open-source framework called ARTEX that turns a plain-language instruction into a full intrusion, run by AI agents at machine speed.
For households and workers, the stakes are direct: the networks the FBI says have been hit include banks, hospitals, government offices and the technology firms that hold their money, their medical records and their personal data. The alert cites one campaign against South Korean financial institutions in which the personal information of about 67,000 people was stolen, and a breach of the American IT company ConnectWise that opened the Microsoft 365 cloud accounts of more than 2,250 of its customers. A break-in that fast leaves a company almost no time to spot the intruder, lock the doors and warn the people whose details were taken.
The FBI's own words carry the scale of the change. ARTEX, it says, "significantly reduces the time required to chain attack techniques, with some intrusions progressing from initial targeting to privileged access in under three hours." Before, breaking into a corporate network was slow, skilled work that took a team days; now, the bureau says, the framework lets attackers with minimal technical expertise run complex, multi-target operations at the same time, automating everything from reconnaissance through to stealing data.
ARTEX itself is not malware. It is a multi-agent penetration-testing platform written in the Go programming language, first released on GitHub in July 2026 by a developer using the name "Autumn-27", and built to probe a network's defences. It contains no AI model of its own; instead it connects to outside large language models — including Anthropic's Claude, OpenAI's GPT, DeepSeek, GLM, Grok and Kimi — and coordinates specialised agents that turn a high-level objective into technical commands. The FBI says attacker-controlled instances most often run Chinese-developed models, led by the DeepSeek Flash family, but the alert names no state sponsor behind the activity.
The bureau says a variety of malicious actors have used ARTEX since at least August 2026, against organisations in ten sectors: financial services, government, technology, defence, semiconductors, energy, healthcare, manufacturing, telecommunications and legal services. Its findings draw on eleven industry and government sources, including Microsoft, Google's threat-intelligence group, CrowdStrike, Censys and Check Point. Microsoft had already linked ARTEX to Storm-0940, a China-state-aligned group that used compromised ConnectWise credentials to target Microsoft 365 environments, and the FBI's alert ties ARTEX infrastructure to Storm-3358, active as recently as November 6. Amazon's threat-intelligence unit, which issued its own warning the same Friday, recorded two ARTEX-linked campaigns between October 3 and November 5 in which agents scanned exposed AI applications, stole provider keys and pooled them, and said it had moderate confidence the operators were separate China-linked groups — while cautioning that it had not watched the activity on the targeted systems as it happened.
The alert also draws a line the Instagram post itself drew: the FBI has not linked these attacks to the recent airline, technology or banking outages in the United States. The bureau's warning describes no new intrusion of its own — it consolidates earlier reporting — and its under-three-hours figure describes some observed intrusions, not a fixed capability or an average across all attacks. The attacks it cites exploited familiar weaknesses rather than a single new flaw: stolen credentials, exposed applications, SQL injection, missing authentication on programming interfaces, hardcoded passwords, unrotated keys and abused cloud permissions.
What the FBI is asking companies to do now is to shrink the surface an attacker can reach: inventory internet-facing applications, self-service portals, cloud infrastructure and supplier systems; lock sensitive services behind trusted networks or dedicated administrative paths; restrict exposed administration tools; cut back application permissions; scan repositories for leaked secrets; and watch privileged identities closely. It has also asked organisations that report ARTEX-related incidents to record two timings — from the first suspicious activity to detection, and from detection to containment — so the bureau can measure how fast the attacks are moving.
The scale of exposure is already visible. Researchers counted more than 1,600 ARTEX control consoles exposed on the public internet by early November, and the FBI says the true number of deployments is likely far higher once systems behind firewalls and on private networks are counted. The original developer took the GitHub repository down in October, but forks, localised versions, archived copies and a paid edition called ARTEX PRO keep it circulating, and security researchers have warned that unofficial copies should be treated as untrusted after Sophos reported a backdoored Docker image of the tool. The alert is marked TLP:CLEAR, meaning it may be shared without restriction.
The warning lands as American businesses are already counting the cost of the year's outages and breaches, and the FBI's message is that the window to catch an intruder has narrowed to hours. Companies that have not yet mapped what they expose to the internet, or that still run on shared passwords and unrotated keys, are the ones the alert is aimed at — and the bureau's next measure of the threat will be the timing reports it has asked victims to send in.