Remember the classmate who never studied but always sat next to the top student during exams? Now imagine the top student never gets tired, never eats, never sleeps, and can sit a thousand exams at the same time.
That, in a nutshell, is ARTEX. And that is why on 19 November 2026 I was on the main evening news of Star, Greece’s national TV channel, in the report on the cyberattack on the European Commission and on attacks that use artificial intelligence as a “weapon”.
What I said on air, I’ll say again here:
“An average hacker with this tool can multiply himself a thousand, two thousand, three thousand times. A target that would normally take a human 3–4 weeks can now be done in 3–4 hours.”
The full news bulletin is on star.gr (in Greek). My part starts at around 15:27.

So what exactly is ARTEX?
ARTEX is an open-source penetration testing tool. Penetration testing means controlled attacks: a company pays someone to attack it so it can find its holes before the bad guys do. The difference from classic tools is that no human drives it step by step. It is run by AI agents, software that decides on its own what to do next.
It was published on GitHub in July 2026 and within ten weeks had 1,500 to 2,000 “stars” from the community. When it started showing up in attacks, its developer closed the source. Too late. Copies and forks are already out there, and as I said on Star, anyone can download it.
By the way, its console reads “autonomous penetration testing console” in Chinese. If you saw it in a news report, that was it.
On 13 November the FBI issued an urgent FLASH alert. According to it, ARTEX has been used since at least August against organisations in banking, government, defence, energy, healthcare, telecoms and legal services. Researchers counted more than 1,600 ARTEX consoles exposed to the internet, and those are only the ones you can see.
Why do I say “a thousand times” and not “a bit faster”?
Because until now, the most expensive part of an attack was human time.
The hacker had to find targets, scan, try, fail, try again, find passwords and climb to administrator rights. That takes weeks. Most importantly, it takes skill. The average hacker got stuck halfway and moved on to the next, easier target.
Now? AI turns him into a conductor of an orchestra of interns who never sleep. He sets a goal and the agents do the rest, in parallel, across dozens of targets. The FBI cites cases where attackers reached privileged access in under three hours. The cost? In the campaigns studied, the AI model bill ranged from $145 to $1,500. Less than a good laptop.
On OPEN I put it more simply: an attack used to need a human, now AI runs a thousand at once. ARTEX is the first tool that makes this real for anyone.
And so we don’t leave it at theory, here is what has already happened:
- ConnectWise: through a supply-chain attack (you hit the supplier to reach its customers), more than 2,250 customer Microsoft 365 environments were exposed.
- Barracuda: someone gave a cloud access key to an outside AI contractor and forgot to revoke it. The attackers breached the contractor, found the key, and then visited about 10 cloud accounts within 44 hours.
- South Korean banks: personal data of about 67,000 people was exposed.

At Tictac, in the incidents we handle, we already see the change in pace. The time from first entry to encryption is getting shorter. And when the attacker moves in hours, the defender who checks the logs “every Monday” has already lost.
So is ARTEX behind this month’s blackouts?
I want to be very clear here, because I saw a lot of people jump to conclusions.
Between 12 and 17 November there was a run of outages: Microsoft 365, Workday, flights in the US, JPMorgan Chase, Apple, even the NSA’s internal network. In the US, people were emptying supermarket shelves. I discussed it the same day on OPEN’s «10 Pantou», in the article Outages or cyberattacks? What AI changes, where you’ll also find the timeline.
But so far, none of these outages has been linked to ARTEX. The FBI itself said it sees no connection. And the causes announced so far are much more… boring:
- Microsoft spoke of a hidden defect in a July update that “woke up” when load increased.
- Workday of a disaster recovery test during which the backup environment connected to production databases by mistake.
- JPMorgan of a software bug.
- Apple of an “internal update”.
- The NSA of “scheduled maintenance”.
See the common pattern? Changes. Upgrades. Patches.
The other protagonist: the flood of CVEs
AI doesn’t only help hackers attack. It also helps the “good guys” find holes, at a pace we’ve never seen. Every hole found gets a CVE code (Common Vulnerabilities and Exposures, an “ID card” for every known vulnerability). And every CVE needs a patch, a fix.
A few numbers to show the scale:
- In early November, Anthropic’s coordinated disclosure programme listed more than 5,000 vulnerabilities in 680 open-source projects, plus more than 37,000 candidate findings waiting for a human to review them.
- Oracle announced 652 new security patches outside its normal cycle.
- Google’s Project Zero is shortening its disclosure deadlines, precisely because AI speeds up both discovery and exploitation.
So big companies today are living through something like renovating a house while still living in it. You knock down a wall, the plumber comes in, then the electrician, and somewhere in between the power goes out. Nobody attacked you. You just wanted to fix everything at once.
There was also a US report, based on three anonymous sources, that Wall Street executives and the NSA told Congress the outages came from attempts to fix vulnerabilities found by AI. This has not been officially confirmed and I won’t present it as fact. But it fits what we see in practice: the speed of change is a risk in itself.
So the picture, as I see it, is not “hackers took us down”. It’s two things at once: attackers became much faster, and defenders are patching at a pace their systems were not designed to handle.
What do I do now?
If you run a business (especially if you work with suppliers who have access to your systems):
- Patch, but with a plan. Test environment first, then gradually, and always with a way back (rollback). “We push everything on Friday night” is not a strategy.
- Revoke old keys and old access. The Barracuda case started with a key nobody remembered. Take inventory: which partner has access, where, and why.
- Turn on MFA everywhere and check sign-ins in your logs. During outages, scammers send “restore your account” phishing. Don’t bite.
- Measure your response time in hours, not days. If the attacker needs 3 hours and you notice in 3 days, you already know who wins. A 24/7 monitoring service (SOC/MDR) helps here.
- Have a plan for when something you don’t control goes down. If your email, bank or cloud goes down for 24 hours tomorrow, how do you keep working? Backups, alternative communication channels, a little cash in the drawer. Yes, that applies at home too.
Think about it: ARTEX was built to protect companies. The same tool, in the wrong hands, attacks them. AI is neither good nor bad. It is a multiplier. The question is whose capabilities it will multiply first: the attacker’s or yours?
Enjoyed the article? Share it with someone who should read it. Your comments are always welcome!
Sources
- FBI warns ARTEX AI hacking tool is powering global cyberattacks – BleepingComputer
- FBI Warns Artex Tool Automates Cyberattacks at Machine Speed – Let’s Data Science
- Barracuda Breached by Attacker Using Artex AI Pentesting Framework – Let’s Data Science
- The November Outage Cluster: Correlated Fragility and Attribution – Cloud Security Alliance
- Open-source AI tool has been used to automate cyber intrusions: FBI – ABC News