More than 800 deployments of Artex, an autonomous AI penetration testing tool, were accessible online without a password, according to cybersecurity firm Censys. The exposure could give outsiders access to systems capable of probing targets for weaknesses and could also reveal tools and information tied to attacks.
What an exposed Artex system can reveal
Artex is open-source software that automatically tests computer systems for vulnerabilities. It is free to download and run on a server, and it has been used by people who cannot access Anthropic or OpenAI’s most advanced cybersecurity AI models.
Microsoft has described Artex as a system of AI agents that can carry out cyberattacks “from reconnaissance through data exfiltration with little needed human interaction.” In practical terms, the tool can automate stages of an intrusion that would otherwise require people to perform them.
Censys scans the internet for exposed or vulnerable systems. In a blog post, the company said it had identified more than 800 distinct Artex deployments that could be reached from the web without a login password.
Silas Cutler, a principal security researcher at Censys and author of the post, cautioned that not every exposed instance is being used maliciously. Some are operated by authorized testers. Still, it is unclear why so many systems were available online: Artex does not need to be internet-accessible to operate.
Exposure creates risks for both sides
A public deployment may be useful to its operator, but leaving it accessible without a password can also let outsiders inspect or misuse it. Censys’s findings illustrate how a configuration mistake can expose a capable attack tool to anyone who finds it.
The risk extends beyond access to the software itself. Public systems can expose attack tools, including implants: hidden malicious software used to compromise a network and maintain access to it. An exposed system may also contain data about activity carried out with the tool.
That creates a problem for both the system’s operator and potential targets. The operator may lose control over who can inspect or use the deployment, while people and organizations targeted through it could face further attacks. The article’s central warning is that a misconfiguration can make the attacker’s own infrastructure visible as well as increase risk to victims.
Artex has appeared in reported attacks
Security companies have linked Artex to several large-scale cyberattacks. One set of attacks targeted the Korean finance industry in September. CrowdStrike attributed them to a Chinese-speaking hacker who used an Artex instance, partly because the hacker had left an “open directory” hosting the system.
Censys said its data indicated that this instance had been exposed for months. It appeared to hold attack data extending “well beyond South Korean banks,” including information involving payment and gaming companies.
Artex was also used in an attack on IT support company ConnectWise earlier in October. Chinese-linked hackers used the tool to break into customer cloud environments and steal data hosted in Microsoft’s cloud, according to Microsoft. The company said the data was taken so quickly that defenders had “little time to detect and respond.”
Many attacks linked to Artex relied on basic weaknesses, including weak or known passwords and the absence of multi-factor authentication. That detail connects the tool’s advanced automation to familiar security gaps: sophisticated capabilities can still find an opening through simple failures in account protection.
Closing access does not erase existing copies
Artex has announced that it would become closed-source, limiting further releases to paid customers. That change may restrict access to future versions, but it does not remove copies already available elsewhere.
Dozens of copies of Artex’s code remain on code-sharing sites and can still be used in active attacks. For organizations, the reported exposure is a reminder to protect internet-facing systems with access controls and to address weak passwords and missing multi-factor authentication. For operators of testing tools, keeping a deployment private when public access is unnecessary can reduce the chance that others will discover and misuse it.