The FBI has issued a stark warning that hackers are now using an artificial intelligence tool to carry out cyberattacks at machine speed, and that the tool is already being used against organisations in almost every major industry around the world.
The warning came in a FLASH advisory issued on 13 November 2026, numbered FLASH-20261113-001, and sent out through the FBI’s Internet Crime Complaint Center under the TLP:CLEAR marking. The subject of the advisory is a tool called ARTEX, an AI-powered penetration testing framework that, in the wrong hands, can automate almost every stage of a cyberattack, from finding a target to gaining privileged access inside a network.
ARTEX is an open-source framework, built on large language models and written in the Go programming language. In plain terms, it works like this: a person describes a high-level goal in ordinary language, such as finding a way into a particular system, and ARTEX translates that goal into machine commands carried out by specialised AI agents. The human operator barely needs to do anything beyond setting the objective. The AI agents do the rest.
That is what makes ARTEX different from the hacking tools that came before it. Traditional cyberattacks require skill, patience and time. An attacker has to probe a system, find a weakness, figure out how to exploit it, and then move carefully through a network without being noticed. ARTEX compresses that process dramatically. The FBI said: “ARTEX significantly reduces the time required to chain attack techniques, with some intrusions progressing from initial targeting to privileged access in under three hours.”
Three hours from first contact to full control. And the danger does not stop at speed. Because ARTEX does the technical work itself, it lets people with minimal technical expertise carry out complex attacks. An attacker no longer needs to be an expert hacker. The FBI said ARTEX lets actors with minimal technical expertise conduct complex, multi-target operations simultaneously, fundamentally accelerating attack velocity and expanding operational scale. Attacks are getting faster, and more people are capable of launching them.
The tool was not built for criminals. It was released on GitHub in July 2026 by a developer known as Autumn-27, who said it was meant for authorised penetration testing. But by early October, reports of abuse were mounting, and on 8 October 2026 the developer removed the repository. By then, it was too late. Forks, localised variants and archived copies were already circulating, and a commercial version called ARTEX PRO had appeared. Taking down the original did nothing to stop the copies.
Since at least August 2026, the FBI says, a variety of malicious actors have used ARTEX to target organisations globally across financial services, government, technology, defence, semiconductor, energy, healthcare, manufacturing, telecommunications and legal industries. That list covers nearly every sector that holds valuable data or runs critical infrastructure. No one industry is the target. Every industry is the target.
The FBI’s advisory describes two campaigns in particular. In one, financially motivated attackers used ARTEX against South Korean financial institutions and exfiltrated personal information from approximately 67,000 individuals. In the other, a supply-chain compromise affecting ConnectWise Cloud Backup customers resulted in attempted sign-ins to more than 2,250 downstream tenants. A single point of compromise became more than two thousand attempted break-ins, because ARTEX can run many attacks at once.
Amazon’s security researchers found the tool is evolving in alarming ways. Between 3 October and 5 November, Amazon’s MadPot honeypot sensors detected two ARTEX campaigns. The first stole AI API keys at what Amazon described as “industrial scale”, harvesting the credentials that give access to powerful AI models, and then feeding those keys into ARTEX to power more agents. The second generated real-time deepfake videos to trick live selfie identity checks at two cryptocurrency exchanges, and forged ID photos to pass document checks.
If that sounds like a threat that is spreading fast, the numbers back it up. As of early November 2026, researchers had identified over 1,600 publicly exposed ARTEX console instances on the internet. The FBI said the total is likely significantly higher when including systems behind firewalls or in private networks. The internet scanning service Censys identified 1,465 hosts running an exposed ARTEX console as of 4 November.
ARTEX itself has no AI model of its own. It plugs into external large language model providers, including frontier models such as Anthropic Claude and OpenAI. But the FBI said the models most commonly used on attacker-controlled systems are Chinese models, led by the DeepSeek Flash family.
The scale of the problem is growing beyond ARTEX alone. Google Cloud’s threat-intelligence group recorded a 73% surge in agentic threat activity over the past month. Agentic threats are attacks carried out by AI agents that can act on their own, and that 73% jump suggests ARTEX is not an isolated case. It is part of a broader shift in which AI agents are becoming a standard tool for attackers, not an experiment.
So what should organisations do? The FBI’s advice starts with shrinking the target. The bureau urged organisations, especially those operating internet-exposed applications, self-service portals, cloud infrastructure and partner or supplier systems, to shrink their external attack surface, restrict sensitive services to trusted networks, and move enterprise applications behind zero-trust access policies. In plain language, that means locking systems down and trusting nothing by default: every request, from every user and every device, has to prove itself before it gets in.
That advice is worth taking seriously, because the old assumptions about cyberattacks no longer hold. It used to be that a complex attack required a skilled team, a lot of time, and a lot of money. ARTEX removes all three. It lets an attacker with little expertise run sophisticated attacks in hours, not weeks.
The FBI’s warning is not about a single tool. It is about a turning point. AI is no longer just helping attackers write better phishing emails or move faster through familiar steps. It is running the attacks itself, chaining techniques together, adapting in real time, and doing it across many targets at once. The warning is simple: the tools have changed, the speed has changed, and the time to close the gaps is now.