On November 6, 2026, TechCrunch reported that researchers had found hundreds of copies of the Artex AI hacking tool sitting openly on the internet, reachable by anyone without a password. Artex is an open-source autonomous penetration testing framework, and it has been tied to real attacks since its public release in July. The twist is that an attacker's own tooling can be misconfigured too, exposing victims and operators alike.

What Artex is and why it spread so quickly

Artex is a penetration testing framework driven by large language models. According to Censys, it was published on GitHub on July 26, 2026 under the AGPL-3.0 license, written in Go, with a Next.js web console and a PostgreSQL backend. A user simply describes a goal, such as gaining access to an admin panel, and AI agents handle the rest.

Under the hood, a Planner agent turns the goal into attack intents, while Worker agents pick them up, run tools such as nmap and report back. The explore-decide-execute loop repeats until the goal is reached. Artex works with models from Anthropic, OpenAI, or any OpenAI-compatible endpoint.

That convenience explains the traction. TechCrunch notes it is a free option for people who cannot get access to the most advanced cybersecurity models from Anthropic or OpenAI. Microsoft has described Artex as able to run attacks from reconnaissance to data exfiltration with very little human input.

Over 800 open systems: what the Censys numbers say

Censys, which scans the internet for exposed systems, reported that as of November 4, 2026 it saw 1,465 hosts running an exposed Artex console. Around 650 of those sit in two blocks of identically configured hosting, leaving roughly 815 distinct-looking deployments. TechCrunch rounded this to "more than 800", and Censys believes the true global figure is probably higher.

Not every instance is malicious; many likely belong to authorized testers. Still, the Artex console does not need to be internet-facing, so its exposure raises obvious questions about configuration habits.

A few other findings stand out:

  • 155 hosts (10.6%) also run at least one other threat Censys detects, including Cobalt Strike and other AI attack frameworks.
  • 32 open directories on Artex hosts currently expose post-exploitation kits, stagers and implants (hidden malicious software).
  • These exposed systems reveal both the attack tools and data about the targets being hit.

From South Korean banks to ConnectWise: where Artex has been used

This is not a theoretical risk. On October 7, 2026, CrowdStrike reported that an unattributed, likely financially motivated Chinese-speaking actor used Artex against South Korean financial firms from late September to early October. The actor was caught because they left an open directory with configuration files and session histories. Censys says that directory had been exposed since August 18, seven weeks before the report, and the files pointed to targets well beyond banks, including payment and gaming services.

The second case involves ConnectWise, an IT support provider. Microsoft says the Storm-0940 group used Artex to break into customers' cloud environments and steal data so fast that defenders had little time to detect and respond. TechCrunch reports that the incident let Chinese-linked hackers spy on thousands of organizations whose data is hosted in Microsoft's cloud.

What many of these attacks share is mundane: guessing weak or known passwords and a lack of multi-factor authentication. AI did not invent new holes; it exploits old ones faster and at larger scale.

Going closed-source did not pull the tool back

After CrowdStrike's report, the author announced the project would go closed-source with no further releases, and the GitHub repository was taken down. But Censys notes that archived copies, translations and rebrands are still circulating, and TechCrunch says dozens of copies remain on code-sharing sites and can still be used in live attacks.

The lesson is familiar: once an offensive tool is out as open source, removal has little practical effect. Businesses should treat autonomous attack agents as a permanent part of the threat landscape rather than a passing event.

How to defend against AI-driven attacks

Speed is what changes most. When an attacker can test many paths in parallel without sleeping, small gaps are found sooner. Prioritize these steps:

  1. Enforce multi-factor authentication on every admin account, mailbox and cloud console; it is the recurring weak point in Artex-linked cases.
  2. Eliminate weak and leaked passwords, with breached-password checks and lockout on repeated failures.
  3. Audit your attack surface: list everything exposed to the internet, especially admin panels, open directories and internal tools.
  4. Monitor cloud activity for unusual bulk access or large downloads to shorten detection time.
  5. Review supply chain access: IT partners with privileged access are another entry point, as ConnectWise showed.

If your team runs its own pentests, protect the testing tools too: put them behind a VPN or IP allowlist, require authentication and never leave data directories open. That exact mistake is what exposed one attacker.

More security stories worth reading

The pattern across recent weeks is consistent: attacks target people and configuration more than exotic technique. See also how a recruitment trap targeted engineers in the $6.3 billion Binance hack, the DreamHost breach that exposed sessions and customer email, and the US emergency warning on industrial freezer cyberattacks. These posts are in Vietnamese.

Conclusion

The exposure of more than 800 Artex servers is a reminder that security is not just about buying good tools: misconfiguration remains the Achilles' heel of attackers and defenders alike. The practical response is not panic but getting the basics right: MFA, strong passwords, a minimal attack surface and cloud monitoring.

Besoft can help you review your systems and build secure software from the start. Contact the Besoft team for a consultation.

Sources

  • TechCrunch, "Hundreds of Artex AI hacking tools found publicly exposed to the web", November 6, 2026: https://techcrunch.com/2026/11/06/hundreds-of-artex-ai-hacking-tools-found-publicly-exposed-to-the-web/
  • Censys, "ARTEX: Agentic Asbestos", November 5, 2026: https://censys.com/blog/artex-agentic-asbestos/