For two years the security industry has warned that AI would change the economics of attack. On Monday the FBI and AWS said the shift has already happened, and gave it a name: industrial scale. In a joint warning, the two said attackers are now using autonomous AI agents to run cybercrime campaigns against critical organizations around the world, moving faster than human defenders can respond. The same day, the FBI's own flash advisory singled out one tool by name, and that detail is what should keep you up at night.

The tool is open source, and it works in under three hours

The FBI says a open-source AI-powered penetration testing platform called Artex has been used in intrusions across financial services, government, tech, defense, telecom and law firm networks, including a widespread hack of ConnectWise customers and across South Korea's financial industry. Artex is not malware in the traditional sense. It is a legitimate category of software built to probe network defenses, and hackers are connecting it to hosted frontier models from Anthropic and OpenAI, as well as open-source and non-US models, to run autonomous attacks. The FBI's most chilling metric: some Artex-linked intrusions compromise a network in under three hours. The agency concedes the tool can be noisy and trip defensive alerts, but says the speed makes it far harder for humans to keep up. Source: https://techcrunch.com/2026/11/16/fbi-says-artex-ai-open-source-hacking-tool-used-to-breach-networks-in-under-three-hours/

Why "industrial scale" is the right phrase

AWS's involvement in the warning matters as much as the FBI's. Cloud providers see the attack traffic at a scale nobody else does, and when AWS co-signs language like industrial scale, it is describing something measured in its own logs, not a hunch. The pair say the campaigns are global and targeted at critical organizations, which fits the pattern security teams have reported all year: reconnaissance, exploitation and lateral movement that used to take skilled operators weeks now compresses into an afternoon. Source: https://www.theregister.com/cyber-crime/2026/11/16/attackers-are-using-ai-agents-to-automate-cybercrime-at-industrial-scale-fbi-and-aws-warn/5310223

The open source twist nobody wants to say out loud

Here is the awkward part. Artex is open source. So, by the FBI's description, is the pipeline around it. The same week that VulnCheck showed Anthropic's Glasswing producing floods of questionable vulnerability findings, and that United Airlines blocked AI agents from its systems, the FBI is pointing at open-source agent tooling as the engine of industrial-scale crime. The lesson is not that open source is bad. It is that agents are dual-use in a way that pipelines and fuzzers never were, because an agent does not just find the gap, it walks through it. Source: https://techcrunch.com/2026/11/16/fbi-says-artex-ai-open-source-hacking-tool-used-to-breach-networks-in-under-three-hours/

What defenders should actually do

The boring advice is now the only advice. Assume breach speed measured in hours, not weeks, which means detection has to be automated too, because a human reading alerts after lunch is already too late. Segment networks so an agent that gets in cannot ride to everything else. Watch for the noise, Artex intrusions generate reams of traffic, and noisy is a signature you can use. And patch on the days vulnerabilities drop, because the window between disclosure and mass exploitation has collapsed to roughly the time it takes to write a prompt. The defenders' own AI agents had better be faster than the attackers'. That is the race now, and nobody at the FBI seems to think we are winning it.