The FBI warned about a hacking tool on Monday. The tool is called Artex. Artex uses AI agents to break into computer networks. A person gives the agents one goal. The agents then scan, break in, and steal data by themselves.
A coder named Autumn-27 built Artex. It went up on GitHub in July. Autumn-27 pulled it down on Oct. 8, after seeing it get abused. Copies and forks still work today.
The FBI says Artex can move fast. Some break-ins reach deep access in under three hours. One attack hit ConnectWise, a company that manages business software. Hackers reached the Microsoft 365 accounts of 2,250 of its customers.
A separate Artex attack hit South Korean banks. It stole personal data on 68,000 people. Security researchers have since spotted more than 1,600 exposed Artex control panels online.
Other Artex attacks hit 14 countries in all. Targets include banks, hospitals, and government agencies. One campaign, called Vapor Panda, hit more than 200 organizations. It broke into at least nine of them.
TechCrunch posted about the FBI warning on X. That post pulled 38 likes and three replies within hours. The story itself reached 40 points and nine comments on Hacker News.
The FBI's warning carried one blunt line. Defenders who "rely on spotting malicious activity and then responding to it will often arrive" after the attack already ends, the agency wrote.
Artex works with many AI models. It can call on Claude, GPT, or several Chinese models too. Attackers can swap in any AI model that answers fastest. Blocking one model just sends them to the next.
This reach says something plain. Hacking used to take real skill and real time. Artex cuts both down to a few hours. Security teams now race a tool that runs at machine speed, around the clock.