November 19, 2026 | News

ARTEX Didn’t Invent New Attacks. It Ended Our Grace Period.

Ragheb Ghandour
19 November 2026

Over the past few days, journalists, activists and organizations across our region have been asking the same question: should we be afraid of ARTEX? On November 13, the FBI published a flash alert, an urgent warning to organizations, about ARTEX, a hacking tool powered by artificial intelligence (AI). 

Attackers have used it to break into organizations around the world, in some cases reaching administrator-level access, the kind that can let an intruder do almost anything inside a system, in under three hours, if no one was watching.

ARTEX exploits the same weaknesses we have warned about for years, such as forgotten passwords or weak security software, to name a couple, but it does it much faster and for much cheaper. The mistakes we used to get away with for months can now be found and abused in an afternoon. The grace period is over. And the best defense against ARTEX is to fix those mistakes.

From security test to attack tool 

ARTEX was built for security testing: organizations hire experts to try to hack them, with permission, so they can find and fix weaknesses before criminals do. A developer called Autumn-27 published it on GitHub in July 2026 and said it was meant for authorized testing only. It is open source, meaning anyone can download, use and change its code. 

It works like this: the attacker types a goal in plain language, and AI agents search for weak spots, break in, look for higher levels of access and collect data, with little human help. ARTEX has no AI engine of its own. It plugs into large language models like ChatGPT, Claude, and DeepSeek. The FBI says Chinese models, especially DeepSeek, are the most common in attackers’ setups because they are affordable, capable and show few visible safeguards when a request is framed as security testing.

Cybercriminals to state-backed hackers has used ARTEX since at least August. In South Korea, attackers stole the personal data of about 67,000 people. At ConnectWise, an IT company, attackers used its cloud backup services to try to break into more than 2,250 customers’ Microsoft 365 accounts. With Barracuda, a security company, the tool found a password for an outside service that was never cancelled and gathered data that way. 

The developer took the project offline on October 8, but it made little difference. Copies, archived versions and a paid edition called ARTEX PRO are still available. By early November, more than 1,600 ARTEX control dashboards were visible on the open internet. The FBI believes the real number in use is much higher. (FBI; the count comes from the internet scanning firm Censys, per The Record).

What ARTEX is and isn’t

ARTEX breaks into organizations’ networks and online accounts. It is not spyware, at least not the kind that is secretly installed on a phone to read messages, track location or switch on the microphone. 

For individual journalists and activists, the biggest threats have not changed: commercial spyware sold to governments, such as Pegasus; phishing; SIM swaps; and phones seized and searched at checkpoints and borders.

The ways in are old and familiar: weak, leaked or stolen passwords; stolen VPN settings; administrator passwords left inside a website’s code; website forms that let anyone upload files without logging in; and accounts belonging to outside companies that nobody switched off. 

Some of its attacks are noisy, creating bursts of unusual activity that security tools can spot, if someone is watching. 

What actually changed

ARTEX changes the economics of hacking, and economics decides who gets targeted. Work that took a skilled hacker days can now take hours. A small group can attack many organizations at the same time, and people, with little technical expertise, can run sophisticated attacks. 

The agents look for forgotten systems that nobody watches, and for trusted providers. In a supply chain attack, attackers break into a company you trust, such as your IT provider, and use its access to reach you. 

A black market is also forming, with China-linked groups stealing API keys, the digital passwords that let software use a paid account (The Record).  A key left in public code is a free account for someone else, billed to you. 

For years, many small organizations relied on being too small to bother with. That protection is gone.

Why our region should pay attention

A quick review from the SMEX ThreatLab did not observe any active instances in the SWANA region exposed to the internet. Regional organizations should still pay attention, since the weaknesses ARTEX exploits are common here.

This points to a visibility gap. Most breaches in our region are never reported, because organizations fear the damage to their reputation, lack the records to know what happened, or have nowhere safe to report. A missing name on a list often means no one looked. 

But, the weaknesses ARTEX exploits are common: one freelance IT person holding the master keys for years, old campaign websites still online, former staff who can still log in, shared email accounts protected by a password alone, and no record of who logged in or when.  Public institutions face the same risks on a much larger scale, holding citizens’ identity documents, addresses and records.

The rights at stake

We can anticipate panic about AI hacking and that demands to outlaw “hacking tools” or expand cybercrime legislation will happen. Across our region,vaguely drafted laws and regulations are frequently used to detain and prosecute journalists, bloggers and human rights defenders. 

A ban would not stop attackers, especially not those who already have their copies. It would stop the researchers who practice responsible disclosure.. This is why we need legal protection for good-faith security research and safe channels to report flaws. Data holders, institutions, and companies should treat two-step login, a full list of exposed systems, activity records kept long enough to investigate, and clear notifications of breaches as obligations. 

ARTEX runs on AI models from major companies as well as on freely available ones. Providers should detect and cut off this kind of abuse, and publish what they find. But the answer cannot be to restrict open AI models or security research, which defenders in our region depend on more than anyone.

What to do now

While we recognize that these steps won’t fix a systemic problem, they can close some doors that ARTEX might leverage. 

For individuals: turn on two-step login, preferably with an authenticator app or physical security key rather than SMS codes; use a password manager; install updates promptly.

For organizations, start this week:

  • List everything you have online and shut down or update whatever nobody uses
  • Turn on two-step login for every account, starting with administrator and shared accounts
  • Cut off former staff, consultants and vendors, including their API keys and connected apps
  • Hide the control panels from the open internet
  • Remove passwords and keys from code, settings files and shared drives
  • Ask your IT provider which tools they use to reach your systems and which outside apps have administrator rights
  • Turn on audit logs and keep them
  • If something looks wrong, save the records first, because deleting accounts or wiping machines too early can erase the only proof of what happened

Public institutions and companies that hold people’s data should apply the FBI’s recommendations in full, hire independent testers, and publish a safe way for researchers to report flaws. The recommendations can be found here. 

ARTEX is a warning about speed. The organizations it breached were beaten by forgotten passwords, unwatched systems and access nobody revoked, found faster than ever before.

If you are a journalist, activist, human rights defender or member of a marginalized group and you think you have been targeted, contact the SMEX Digital Safety Helpdesk. It is free, and you can reach it by text or voice message on Signal or WhatsApp at +961 81 633 133. 

Further reading below:

ARTEX Didn’t Invent New Attacks. It Ended Our Grace Period.

Over the past few days, journalists, activists and organizations across our region have been asking me the same question: should we be afraid of ARTEX?

On November 13, the FBI published a flash alert about ARTEX, a hacking tool powered by artificial intelligence (AI). Attackers have used it to break into organizations around the world. In some cases, they reached administrator-level access, the kind that lets an intruder do almost anything inside a system, in under three hours. The headlines followed, and so did the panic.

My answer, as a technologist who spends his days investigating hacks against civil society is the following. ARTEX did not invent new ways to attack. It exploits the same weaknesses we have warned about for years, only much faster and much cheaper. What it ended is our grace period. The mistakes we used to get away with for months can now be found and exploited in an afternoon. The best defense against ARTEX is to fix those mistakes. 

What the FBI actually said

The FBI alert, titled “Threat Overview of ARTEX Autonomous Penetration Testing Framework,” was released on November 13, 2026 (FLASH-20261113-001). It is marked TLP:CLEAR, a label meaning anyone may share it freely.

To understand ARTEX, start with understanding penetration testing, a security exercise where organizations get security experts to attack their own systems, with permission, so they can find and fix weaknesses before criminals do. ARTEX was built to automate that job. A developer using the name Autumn-27 published it in July 2026 on GitHub, the website where programmers share code, and said it was meant for authorized testing only.

ARTEX is open source, meaning anyone could download, use and change its code. It works through AI agents, AI programs that do not just answer questions like a chatbot, but take actions on their own, one step after another. The attacker types a goal in plain language, such as “get into this company’s email.” The agents then search for weak spots, break in, look for higher levels of access and collect data, with little human help.

ARTEX has no AI engine of its own. It connects to large language models, the AI systems behind chatbots, made by companies including OpenAI (GPT), Anthropic (Claude), DeepSeek and others. The FBI says Chinese models are the most common in the setups attackers run, especially DeepSeek, because they are cheap, capable and show few visible safeguards when a request is framed as security testing.

According to the FBI, several groups, from cybercriminals to state-backed hackers, have used ARTEX since at least August 2026. The known cases include:

  • Banks and financial companies in South Korea, where attackers stole the personal data of about 67,000 people (FBI).
  • ConnectWise, a US company whose software IT providers use to manage their clients’ systems. Attackers abused its cloud backup service to try to break into more than 2,250 of its customers’ Microsoft 365 accounts, which hold their email and files (FBI). Microsoft attributes this attack to Storm-0940, a group it describes as aligned with the Chinese state (OODA Loop).
  • Barracuda, a security company, where the tool found a password for an outside service that was never cancelled after the business relationship ended (TechCrunch).

The developer took the project offline on October 8 and it made little difference. Copies that others had made and modified, archived versions and a paid edition called ARTEX PRO are still available. By early November, more than 1,600 ARTEX control dashboards were visible on the open internet, and the FBI believes the real number in use is much higher (FBI; the count comes from the internet scanning firm Censys, per The Record). A quick review from the SMEX ThreatLab did not observe any active instances or ARTEX in the SWANA region that are exposed to the internet. 

What ARTEX is not

Let me be clear about what this tool does not do, because fear travels faster than facts.

ARTEX is not spyware. Spyware is software secretly installed on a phone to read messages, track location or switch on the microphone. ARTEX does none of that. It is a tool for breaking into organizations’ networks and online accounts.

For individual journalists and activists in our region, the biggest threats have not changed:

  • commercial spyware sold to governments, such as Pegasus, which researchers found used widely against journalists and civil society in Jordan (GCHR);
  • phishing, fake messages and websites designed to steal your passwords;
  • SIM swaps, where someone convinces your phone company to move your number to their SIM card and receives your login codes;
  • phones seized and searched at checkpoints and borders.

ARTEX does not use secret new weaknesses. Security experts track publicly known software flaws in a global catalogue, where each flaw gets an ID called a CVE. The FBI alert names no specific CVE (OODA Loop). The ways that it lists are old and familiar (FBI):

  • weak, leaked or stolen passwords;
  • stolen VPN settings that would enable remote connection to an office network;
  • administrator passwords left inside a website’s code, where anyone who looks can find them;
  • website forms that let anyone upload files without logging in;
  • common website flaws such as SQL injection, a trick that makes a website’s database hand over its data;
  • accounts belonging to outside companies that nobody switched off.

In South Korea, reporting points to missing protections such as two-step login (TechCrunch). At Barracuda, it was a forgotten password.

ARTEX is not invisible: The FBI notes that its attacks are noisy: they create bursts of unusual activity that security tools can detect (TechCrunch). The “three hours” headline describes the fastest cases, not every attack. That figure is real and alarming, but it describes what happens to organizations that leave their doors open.

What actually changed

If the weaknesses are old, why does this matter? Because ARTEX changes the economics of hacking, and economics decides who gets targeted. The FBI’s own list of risks shows how (FBI):

  • Speed. Steps that took a skilled hacker days can now take hours, leaving defenders very little time to notice and react.
  • Scale. Small groups can now attack many organizations at the same time, stepping in only when a decision is needed.
  • Skill. People with little technical expertise can now run sophisticated attacks.
  • Forgotten systems. The agents look for the systems on the edges that nobody watches: partner portals, the servers behind mobile apps, old services still running, and connections to outside software.
  • Trusted providers. This is what experts call a supply chain attack. Instead of attacking you directly, attackers break into a company you trust, such as your IT provider or a software vendor, and use its access to reach you. Whatever access your IT provider has to your systems is now part of your risk.

A black market is also forming around these tools. Amazon reported that China-linked groups are stealing API keys, the secret codes that let software use paid AI services, in an operation where stolen keys pay for further theft (The Record). A key left in public code is no longer a small mistake: it is a free account for someone else, billed to you.

For years, many small organizations relied, without saying so, on being too small to bother with. That protection is gone. When attacking costs almost nothing, everyone is worth attacking.

Why our region should pay attention

None of the victims named so far are in West Asia or North Africa. That is not reassurance. It is a visibility gap.

The countries in the headlines are places where security companies monitor networks closely and businesses are often required to disclose breaches. In my experience, most breaches in our region are never reported. Organizations fear the damage to their reputation, lack the records to know what happened, or have nowhere safe to report. A missing name on a list often means no one looked, not that nothing happened.

Meanwhile, the weaknesses ARTEX exploits are common here. In my work with civil society organizations, I regularly find:

  • one freelance IT person or small web hosting company holding the master keys to everything, for years;
  • old project and campaign websites still online, running software nobody has updated since the grant ended;
  • staff and consultants who left long ago but can still log in;
  • shared email accounts protected by a password alone;
  • no activity records (logs), or records nobody keeps.

Public institutions face the same risks on a much larger scale. Government online services across the region increasingly hold citizens’ identity documents, addresses and records, and government is one of the sectors the FBI says has been hit (The Record). When those platforms are breached, the people harmed are citizens who never had a choice about handing over their data.

The rights questions behind the headlines

As a digital rights advocate, I worry less about ARTEX itself than about how governments may respond to it.

Protect cybersecurity defenders: Like many tools standardly utilized by security experts, ARTEX originally served as an authorized security testing mechanism. We can anticipate reactive demands to outlaw “hacking tools” or expand cybercrime legislation. Throughout our region, such vaguely drafted statutes are frequently weaponized to detain and prosecute journalists, bloggers, and human rights defenders for online expression.

A ban would not stop attackers, who already have the copies. It would stop the researchers who find flaws and report them privately so they can be fixed, a practice known as responsible disclosure. What we need instead is legal protection for good-faith security research, and clear, safe channels to report flaws.

Hold data holders accountable: If a ministry, a bank or a telecom company holds our data, basic protections should be obligations, not favors: two-step login, a full list of every system data holders expose to the internet, activity records kept long enough to investigate, and clear informative noifications to  people when a breach happens. Barracuda, a security company, could not rule out that customer data had been accessed because its records were not detailed enough (TechCrunch). How many public institutions in our region could answer that question?

Refuse fear as a pretext: Panic about AI hacking can easily become the excuse for more surveillance, more censorship and more state control over the internet. SMEX research on digital authoritarianism in Egypt, Iraq and Tunisia carries a title that says it plainly: “Fear Is the Power That Can Do Anything.” Security that protects people looks like updated systems and accountable institutions, not wider powers to monitor citizens.

Ask more of AI companies, without shutting down open research: ARTEX runs on AI models from major companies as well as on freely available ones (FBI). Providers should detect and cut off this kind of abuse, and publish what they find. But the answer cannot be to restrict open AI models or security research, which defenders in our region depend on more than anyone.

What to do now

If you are an individual:

  • Turn on two-step login (also called two-factor or multi-factor authentication) for your email, messaging and social media. After your password, the service asks for a second proof, such as a code from an app on your phone. Use an authenticator app or a physical security key rather than SMS codes where you can, because SMS codes can be stolen through a SIM swap.
  • Use a password manager, an app that creates and remembers a different strong password for every account, and stop reusing passwords.
  • Install updates on your phone, computer and apps promptly. Many updates fix security flaws.

If you run an organization, start this week:

  1. List everything you have online: websites, old project pages, extra web addresses such as old.yourorg.org, staff portals and cloud accounts. The FBI’s first recommendation is to know this list and shrink it (FBI).
  2. Shut down or update whatever nobody uses or maintains.
  3. Turn on two-step login for every account, starting with administrator accounts and shared accounts.
  4. Cut off people who have left: former staff, consultants and vendors. Remove their accounts, their API keys and any apps they connected to your email or files.
  5. Hide the control panels. The dashboards used to manage your website, hosting and databases should not be reachable by anyone on the internet. Limit them to trusted networks or a secure connection (FBI).
  6. Get passwords out of documents and code. Remove passwords and keys stored in code, settings files and shared drives, and change any that were exposed.
  7. Ask your IT provider two questions: which tools do you use to reach our systems, and which outside apps have administrator rights in our Microsoft 365 or Google Workspace?
  8. Turn on activity records (audit logs) for email and cloud accounts, and keep them. They show who logged in, when, and what they did.
  9. If something looks wrong, save the records first. The FBI advises preserving evidence before removing an intruder’s access (FBI). Deleting accounts or wiping machines too early can erase the only proof of what happened.

If you run a public institution or a company that holds people’s data: apply the FBI’s recommendations in full, hire independent testers to probe your systems before attackers do, and publish a safe way for researchers to report flaws to you.

Attacks got cheaper, protecting people didn’t.

ARTEX is a warning about speed, not about magic. The organizations it breached were not beaten by a superhuman AI. They were beaten by forgotten passwords, unwatched systems and access nobody revoked, found faster than ever before.

That is why the unglamorous work matters: keeping systems updated, training teams, investigating hacks and reporting flaws before attackers find them. At SMEX, our Threat Lab investigates the digital threats facing people in our region, and our Digital Safety Helpdesk responds when someone is under attack. This is the work funders, institutions and governments need to support before the next breach makes headlines, not only after.

If you are a journalist, activist, human rights defender or member of a marginalized group and you think you have been targeted, contact the SMEX Digital Safety Helpdesk. It is free, and you can reach it by text or voice message on Signal or WhatsApp at +961 81 633 133.

Key terms, in plain language

Term What it means
Administrator access The highest level of control over a system, enough to read, change or delete almost anything. Also called privileged access.
AI agent An AI program that takes actions by itself, step by step, instead of only answering questions.
API key A secret code that lets one piece of software use another service, often a paid one.
Large language model The AI engine behind chatbots such as ChatGPT or Claude.
Logs (audit logs) Records a system keeps of who logged in, when, and what they did.
Open source Software whose code anyone can download, use and change.
Penetration testing A simulated attack that an organization authorizes, to find its own weaknesses before criminals do.
Phishing A fake message or website designed to trick you into handing over a password or installing something harmful.
Responsible disclosure Reporting a security flaw privately to its owner so it can be fixed before it is made public.
SIM swap Taking over your phone number by convincing your mobile operator to move it to a SIM card the attacker holds.
Spyware Software secretly installed on a device to watch the person using it.
Supply chain attack Breaking into a trusted supplier, such as an IT provider, to reach its clients.
TLP:CLEAR A label on a security report meaning it can be shared publicly.
Two-step login (MFA) A second check after your password, such as a code from an app or a physical security key.
VPN A secure connection that lets staff reach an office network from outside.
Vulnerability, CVE A flaw in software that attackers can exploit. A CVE is the ID number given to a publicly known flaw.

Sources and further reading

Recent Updates

The latest updates on digital rights from West Asia and North Africa

02 Nov 2026

Lebanon’s Fight for Digital Sovereignty, Reimagined

30 Oct 2026

Internet Governance Forum 2026: Lebanon Discusses Digital Recovery as Reforms Remain Stalled

30 Oct 2026

Comments on Lebanon’s National Artificial Intelligence Strategy 2032

Stay Informed.

Subscribe to our monthly newsletter.