News: We're testing our own network the way attackers do
Attackers are now using AI tools that can find a weak spot on a network and break in within a few hours, with very little human effort. On November 13 the FBI issued a warning about one of these tools. We've responded by running the same kind of testing against our own network, and the internet addresses we provide to you, so we find the gaps first.
What's happening
Chaining together the steps of a break-in used to take a skilled person and a lot of slow, manual work. That's changing quickly. AI models can now plan an attack, run the hacking tools, read the results and decide on the next step, over and over, with very little human input.
- The FBI warning. The FBI's FLASH alert FLASH-20261113-001 describes ARTEX, a free, open-source AI attack tool. Since August 2026 attackers have used it against financial firms, governments, technology firms, hospitals, manufacturers, law firms, and phone and internet companies. Some break-ins went from first contact to privileged access, the kind of access an administrator has, in under three hours.
- Real damage. The FBI links the tool to the theft of personal information on about 67,000 people from South Korean financial institutions, and to a breach at a cloud backup provider that gave attackers access to more than 2,250 of its customers' cloud environments.
- It isn't going away. The original author pulled the tool down in October, but copies keep circulating, and researchers have found more than 1,600 copies of its control panel running openly on the internet. TechCrunch and Cisco Talos both cover how quickly criminal groups picked it up.
- It's bigger than one tool. In November 2025, Anthropic reported a state-sponsored spying campaign against about 30 organizations in which AI did an estimated 80 to 90 percent of the work. Google's threat intelligence team has documented malware that calls AI models while it runs to write new code and hide itself, and AI hacking kits advertised to less experienced criminals.
The practical effect is speed and scale. A small group can now probe thousands of networks at once, and anything left exposed gets found much sooner than it used to.
What these tools look for
According to the FBI, the AI isn't inventing new kinds of attacks. It works through the ordinary mistakes people have always made, just much faster and without getting tired. It goes looking for:
- Admin pages for routers, firewalls, servers, phone systems and backup systems that can be reached from the internet
- Default or weak passwords, and logins that never lock out after repeated failures
- Passwords and keys accidentally left in website code
- Software that's out of date and has known security flaws
- Web apps and APIs that answer requests without asking who's calling
- Forgotten systems: old test servers, backup files, and side portals nobody watches as closely as the main website
What we're doing about it
We built our testing directly from the FBI's list, and we run it from a computer outside our network, so we see what an outside attacker would see. That computer sits in Las Vegas at Fiberhub, a data center whose security and network know-how made them a great match for this work. Thanks to Rob Tyree and his team for the help. Our testing covers three areas:
- Our core systems. Our client portal, network equipment, DNS, phone platform and the other systems that keep your service running.
- The services we host for you. Website hosting, email and hosted phone systems.
- Every internet address we provide. That includes the addresses assigned to your equipment.
Across those areas we:
- Look for services that shouldn't be reachable from the internet, like remote desktop, database ports, device admin pages and old protocols.
- Check for settings that can be abused in attacks on others, like open DNS resolvers.
- Read the software version each reachable service announces, and compare it against published lists of known vulnerabilities, including the US government's list of flaws that are actively being exploited.
- On the services we run for you, like webmail, website hosting and our client portal, make sure repeated failed logins get blocked. These tests use made-up usernames, never real accounts.
- Scan our own web applications for leaked passwords or keys and for known security flaws.
We've also compared our network records against the addresses the FBI tied to this activity, and we're working through the FBI's other recommendations for our own systems, starting with keeping management interfaces off the public internet.
We also work with the Canadian Centre for Cyber Security, the federal government's cyber security agency. Its National Cyber Threat Notification System sends us reports almost every day about addresses on our network that look exposed, vulnerable or infected. We pass each one on to the customer it affects, and our own scan now double-checks every report, so we can tell whether a problem is still there or has been fixed.
AI helps us sort through the results quickly, so real problems get fixed first. It's the same speed advantage the attackers have, used on the defending side.
What we don't do
We only look. We don't log in to your devices or websites, try passwords on them, try to break in, or change anything on your equipment. And if you'd rather we didn't look at all, you can opt out (see below).
Opting out
If you'd rather we didn't include the internet addresses we've assigned to you, email support@incentre.com or open a support ticket with your account name and the addresses. We'll take them out of our testing before the next scan, and you can opt back in the same way at any time.
Opting out covers the equipment on addresses we've assigned to you. It doesn't apply to the systems we run on your behalf, like our web hosting, email and hosted phone servers, because keeping those safe is our job. Keep in mind that if you opt out, we won't be able to warn you about an exposure we would otherwise have spotted.
What you can do
- Keep the firmware on your router, firewall and phone system up to date. Out-of-date software is one of the first things these tools check.
- Don't leave admin pages, remote desktop or file sharing open to the internet. If you need remote access, use a VPN.
- Change every default password, use a different password for each system, and turn on multi-factor authentication wherever it's offered.
- If you're not sure what your network shows to the internet, ask us. We're happy to go over what we see on your addresses.
What you might notice
If you run a firewall or intrusion detection, you may see connection attempts from 172.93.1.19 (scanner.incentre.net). That's us, and visiting that address explains what it's doing. The tests run at a low rate and won't affect your service.
If our scan or the Cyber Centre flags something on an address we've assigned to you, we'll email you from support@incentre.com with what we saw and how to close it. So these stay useful rather than noisy, we send at most one of these emails a week.
Further reading
- FBI FLASH-20261113-001: Threat Overview of ARTEX Autonomous Penetration Testing Framework (PDF)
- TechCrunch: FBI says Artex AI open source hacking tool used to breach networks in "under three hours"
- Cisco Talos: ARTEX and beyond, AI-enabled penetration testing platforms
- Canadian Centre for Cyber Security: National Cyber Threat Notification System
- Anthropic: Disrupting the first reported AI-orchestrated cyber espionage campaign
- Google Threat Intelligence Group: Advances in threat actor usage of AI tools