Security researchers indicate that a new era of AI-driven cybercrime has arrived with the widespread adoption of Artex, an open-source artificial intelligence framework that automates penetration testing. Since late August, this tool has been used by cybercriminals to target banks, government agencies, and corporations globally, enabling attacks to be completed in hours rather than days.
Artex, initially published on GitHub in July by a developer known as Autumn-27, is not an AI model itself but a guide that directs external AI models to execute a series of hacking tasks. This “point-and-shoot” approach allows users to issue natural language commands, after which multiple AI agents collaborate to scan targets, breach networks, and extract data. The tool’s simplicity lowers the skill barrier, making it accessible to unskilled actors who lack traditional cybersecurity expertise.
The FBI warned last week that various malicious actors, including Chinese-speaking groups, have used Artex to conduct campaigns across sectors such as finance, defense, and healthcare. In October, a China-linked group known as Vapor Panda used the framework to target over 200 organizations, while Microsoft identified a separate group using Artex to exploit a supply chain compromise, gaining privileged access in under three hours. Although the original developer removed the code from GitHub to distance themselves from malicious use, copies and similar tools remain prevalent in hacker communities.
Experts note that while these AI-driven attacks are highly efficient, they are currently not stealthy. For example, the platform Dosu reported receiving 80,000 requests from an Artex-based attack in 90 minutes, creating significant network noise that is detectable by standard monitoring systems. Security analysts from firms like Censys and GitGuardian have documented dozens of similar AI attack frameworks, many of which have been left exposed on the internet, offering defenders valuable insights into criminal tactics. Despite the sophistication of some organized threats, researchers also observed instances of juvenile activity, such as a password spraying attack on a university, suggesting a mixed landscape of criminal intent behind the tool’s usage.
It is fascinating how a simple framework can empower state actors like the China linked groups mentioned.
I hope we see better monitoring tools soon because this level of network noise is becoming a constant problem.
The fact that natural language commands automate the entire breach process just changes everything for criminal groups.
Removing it from GitHub is useless now because the code has already spread across every dark web forum.
Hate to admit it, but the password spraying on a university was probably just bored kids testing limits.
This feels like a wild west era where supply chain compromises happen in under three hours.
Do you think banks will start using AI agents to defend against these automated hacking frameworks effectively?
Eighty thousand requests in ninety minutes is not stealthy at all, so defenders can still catch these attacks easily.
I am surprised the original developer did not anticipate this misuse when they uploaded the open source code.
It is scary how much damage one unskilled teenager can do now with natural language commands.