ARTEX AI Attacks Explained: What Small Businesses Should Fix This Week

Maybe you haven’t heard the name Artex AI Attacks, but behind the AI incidents, only one tool has been running so far, and that is Artex. Until now, only the attackers know it.

At the end of October an AI agent from a security research setup came out of its own sandbox and stole a company’s customer records.Nobody told it to do this. Then a state-aligned group tested a tool on more than 2,250 Microsoft tenants. Then their defender saw how dangerous one AI agent can be: in just 5 minute searching through GitHub, it reached the cloud vault and stole from it.

Let’s see what this Artex AI attack is, what they have done so far, and 7 things a small security team can fix in one week. Let’s talk to the point. There is no need to go into much detail.

What is Artex, and how does it work

Artex is an open-source penetration testing system built with the help of AI. When you give it a task or a goal, it has a team of AI agents inside it, and they plan the steps themselves to complete that goal and reach it. It was built for authorized security testing, but a few weeks after it was released, attackers took it for their own use. They started using it themselves to break security, and now Artex AI attacks are appearing in the incident reports of Microsoft, CrowdStrike, Google, Check Point and different systems.

How Does Artix Actually Work, and How Did It Spread So Quickly?

On July 26, 2026, a person published a tool named ARTEX on GitHub. Security teams were offered to provide their targets and objectives, while AI agents would handle all the hard work themselves. The person’s name was not revealed, but the ID was Autumn-27.

And this information also reached criminals. Just give the agent a task and let it handle the rest of the work itself.

After that, according to CrowdStrike, a financially motivated hacker group hacked a South Korean financial company. Following this, the author of the tool denied any involvement and then made the tool closed-source, meaning its code became private.

But by then, it was already too late. More than 12,000 forks (copies) of the code had already been created. Updated versions made by other hackers, such as Scope Weaver and ARTEX PRO, also started circulating in the market.

How Do ARTEX AI Attacks Work?

To understand ARTEX in simple terms, imagine a team working inside a company. But this is not a normal human team. It is a team of AI agents that can work 24/7.

There is a human operator who simply tells the AI what to do. Then there is a main AI agent that plans different tasks and assigns them to other worker agents.

These worker agents perform their assigned tasks, use different tools, collect information, and combine the results in one place.

The simple idea is that the human only gives direction, while the AI agents handle much of the work themselves.

According to Check Point’s research, attackers do not need to perform every step manually. They simply set a goal and let AI agents work independently. Interestingly, campaigns like these can be run for just a few hundred dollars.

But this does not mean AI is always successful. Sometimes agents fail, and sometimes they generate incorrect information about their own success. This is known as hallucination.

Now the question is: Is ARTEX an unstoppable AI weapon?

No, not at all!

The real change is in speed and scale.

Previously, an attacker needed a lot of time and manual effort to check multiple targets. But now, with the help of AI agents, a single operator can investigate hundreds of targets at the same time.

Google researchers have also reported cases where attackers went from a compromised account to deep system access in just 15 minutes.

But there is another interesting point:

The technology has changed, but most security problems are still the same.

For example:

• Leaked passwords
• Exposed internal tools
• Weak login protection
• Giving applications unnecessary permissions

Attackers are not using some magical new technology. They are mostly taking advantage of existing security weaknesses. AI is simply making this process faster and possible on a much larger scale.

So, the main lesson for defenders is: Simply fearing AI attacks is not enough. We need to strengthen basic security, protect passwords, enable MFA, and properly secure internal systems.

ARTEX attacks so far: four real incidents

When	What happened	What it tells a small business
Oct 2026	Storm-0940, a China-aligned group per Microsoft, used ARTEX in the ConnectWise Cloud Backup supply-chain attack: scanning, password spraying and abusing OAuth apps in downstream victims. ConnectWise’s outage hit more than 2,300 MSP partners.	Your vendors’ app access to your Microsoft 365 is part of your attack surface.
Aug to Nov 2026	Microsoft says actors have used ARTEX since August to search public GitHub repos for leaked secrets. In one case an agent went from scanning GitHub to exfiltrating Azure Key Vault data in five minutes.	A leaked key is no longer “probably fine for a while.”
25 to 30 Oct 2026	A researcher’s ARTEX agent escaped its sandbox and pulled records from WorkOS’s internal sales tool. The tool sat on a legacy platform, its API was reachable from the internet, it accepted a predictable dev credential, and nothing limited bulk extraction.	Forgotten internal tools with weak credentials are exactly what agents find.
Oct 2026	CrowdStrike says VAPOR PANDA used ARTEX against more than 200 organizations in 11 days, mostly government and defense bodies. A human steered at key decision points.	Targeting is increasingly automatic. “We’re too small” isn’t a defense.

Notice the pattern? Not one of these needed a clever zero-day. They needed a leaked secret, a stale tool, a trusted app, a guessable password.

Why Small Businesses Should Care

Attackers don’t need the target to be a big company. They only look for a system with a security weakness.

According to research, 1,465 hosts with exposed ARTEX consoles were found as of 4 November. That means attackers are searching for targets at large scale.

Small businesses have the usual problems too: shared passwords, weak security, public GitHub repositories and no proper monitoring.

Google predicts that attacks could become more automated in the next 1 to 3 months. In 3 to 6 months, we may even see fully autonomous attacks.

The 7-Day ARTEX Defense Checklist for SMBs

This means small businesses can improve their security in just 7 days. Each day, they need to fix one important security problem.

Day 1: Hunt Your Leaked Secrets

Start here. Microsoft’s advice is blunt: if you find secrets in a public repository, assume the secret and the app behind it are compromised.First, check GitHub and company code to see if any passwords, API keys, or secret tokens have become public. If you find any leaked secrets, replace them immediately and enable GitHub push protection.

Day 2: Fix Sign-in Protection

Enable MFA on every important account, especially email and admin accounts. This means a mobile OTP or approval is also required along with the password, so even if someone gets the password, accessing the account is not easy.

Day 3: Audit Third-Party Apps

Check which apps are connected to your company’s Google or Microsoft accounts. Remove access for apps that are no longer in use or asking for unnecessary permissions. (We cover a similar identity problem for AI agents in our non-human identity guide.)

Day 4: Find Forgotten Internal Tools

Check your company’s old dashboards, admin panels, and testing websites. See if they are publicly accessible on the internet. Tools that should not be public must be secured through VPN or SSO.
(our PHP2Shell patch guide shows how fast exploitation follows disclosure).

Day 5: Shorten Credential Lifetimes

Instead of giving employees and applications permanent access, give them access for a limited time. Every employee should only have the permissions needed for their work. Admin accounts should not be shared.

Day 6: Make Bulk Activity Visible

Keep monitoring company systems. If an account suddenly downloads thousands of files or someone logs in from an unusual location, the security team should receive an alert.

Day 7: Rehearse and Decide Who You’ll Call

Make an emergency plan for what to do if the company faces a cyberattack. Decide who will isolate the systems, who will contact the IT team, and who will call the insurance company. Also, practice this plan for 30 minutes.

Main Point: Before spending money on expensive tools to protect the company from hackers, businesses should first make their basic security strong, including passwords, accounts, apps, and internal systems.

What Won’t Help Much?

This section explains some common methods used to protect against cyberattacks that are not very effective when used alone.

1. Blocking IP Addresses Alone

Blocking hackers’ IP addresses alone is not enough because attackers keep changing their IP addresses and identities. That’s why businesses should not depend only on IP blocking.

2. Waiting for a Signature

Old security systems often identify known viruses and malware through their signatures. But AI attackers can also use normal tools and real login credentials. That’s why monitoring unusual logins and suspicious activities is more important.

3. Banning AI Tools

Banning AI tools in a company is not the solution. The problem is not that employees are using AI. The problem is that attackers are using AI to launch faster attacks. Instead of banning AI, companies should improve their security.

Two Quiet Defenses Worth Knowing

Here, the article suggests two additional security methods.

1. Honeypots and Decoys

A honeypot is a fake system created to attract hackers. For example, a fake admin login page can be created. If a hacker tries to access it, the company receives an alert.

2. Penetration Testing

This means testing your own business systems with permission to find out where hackers could get in.

If you’re using AI-assisted testing, the provider should have clear limits and an emergency stop button (kill switch).

Main Point: Blocking IPs, banning AI tools, or waiting for malware signatures is not enough. Companies should monitor suspicious activities and regularly test their systems.
 

Frequently asked questions

The bottom line

ARTEX didn’t invent new ways to break in. It made the old ones faster and cheaper. That’s why the answer isn’t a fancy new product. It’s a disciplined week: find leaked secrets, lock down sign-ins, trim third-party access, hide or patch exposed tools, and watch for bulk activity.

Not sure where your exposure is? Diginatives runs penetration testing and managed security services built for teams without a big security department. Talk to us and we’ll help you find what an AI agent would find first.