Security firm Censys has identified more than 800 distinct deployments of Artex, an open-source Chinese-made autonomous penetration-testing tool, that anyone can reach on the internet without a password. TechCrunch reported the finding on November 6; the tool has been linked to several large cyberattacks since its public release in July.
What Censys found
Artex is software that automatically probes computer systems for weaknesses. Microsoft has described it as a system of AI agents able to conduct attacks “from reconnaissance through data exfiltration with little needed human interaction.” Because it is free and open source, it is popular with people who cannot get access to the most advanced cyber models from Anthropic or OpenAI.
Silas Cutler, a principal security researcher at Censys, wrote in a blog post that more than 800 Artex instances were open to the web with no login. Not every instance is malicious, since many belong to authorized testers, but Artex does not need to be internet-facing, so it is unclear why so many are. The exposed systems also reveal attack tooling, including implants, which are hidden programs used to keep access to a victim’s network.
Attacks already linked to Artex
TechCrunch lists several incidents. CrowdStrike attributed attacks on the Korean finance industry in September to a Chinese-speaking hacker who used an Artex instance, partly because that person left an open directory hosting it. Censys says the instance had been exposed for months and held attack data reaching “well beyond South Korean banks,” including payment and gaming companies.
Artex was also used in an attack on IT support provider ConnectWise earlier in October, which let Chinese-linked hackers spy on thousands of organizations whose data sits in Microsoft’s cloud. Microsoft said the attackers moved so fast that defenders had little time to respond. Many of these attacks relied on basic failures, such as weak or known passwords and missing multi-factor authentication.
The project has since announced it will go closed-source, which limits further releases to paying customers. Dozens of copies of its code remain on code-sharing sites and can still be used.
How open and closed cyber models compare
A WIRED report on November 6 adds context. The testing company Vals ran models through simulated corporate networks and found OpenAI’s gated cyber model Daybreak completed 81% of tasks on average, while the open-weight DeepSeek completed 73%. Vals put the token cost of an attack at $17 to $46 for GPT-6 Sol versus $2 to $3 for DeepSeek. These are figures from Vals as reported by WIRED, and we have not seen the underlying data.
The gap matters because gated release was meant to buy defenders time. Our earlier coverage of OpenAI’s GPT-6 Cyber and Daybreak, Mistral’s open-weight Large 4 aimed at defenders and our quick take on gated versus open cyber models traces that debate.
What defenders should take from it
For organizations, the incidents point to unglamorous controls: unique strong passwords, multi-factor authentication, and patching, because AI-driven tools tend to exploit the simplest weakness first. For security teams running their own testing tools, the lesson is not to expose them to the internet: the exposed Artex servers harmed their operators as well as potential targets.
What to watch next
- Whether Censys’s count of exposed Artex deployments falls after the disclosure.
- How far the closed-source move limits Artex’s spread, given copies already online.
- Further vendor reports attributing intrusions to Artex.
- New Vals or other evaluations of open-weight models on multi-step network tasks.
- Policy responses on access controls for cyber-capable models.
Key takeaways
- Censys counted more than 800 Artex deployments reachable without a password.
- Artex is free, open source and has been tied to attacks on Korean banks and ConnectWise.
- Many attacks used weak passwords and missing multi-factor authentication.
- Vals results reported by WIRED suggest open-weight models are close to gated ones on cyber tasks.
Frequently asked questions
What is Artex?
Artex is an open-source, Chinese-made autonomous penetration-testing tool that uses AI agents to probe systems for weaknesses. It was released publicly in July.
Why are the exposed Artex servers a problem?
Anyone can reach them without a password, so outsiders could launch attacks through them or view attack tooling and stolen data they hold.
Has Artex gone closed-source?
Yes, the project announced it will limit further releases to paying customers, but dozens of copies of earlier code remain available online.
