A spy agent on autopilot
According to CloudSek, the hackers began using Hermes for reconnaissance around mid-August and, roughly two weeks later, broke into Tajikistan’s electronic document system. The AI agent then wrote a custom program to download the files, sort them and split them into chunks so they could be siphoned off more efficiently.
For the researchers, the significance lies less in the haul than in the method. Reuters has previously reported on cybercriminals assigning break-ins to autonomous agents; CloudSek said this discovery showed that more sophisticated espionage groups are now doing the same, delegating the tedious technical work of an intrusion to software that never tires.
“Instead of typing hacking commands by hand, the attacker told Hermes what to do, and Hermes planned the attacks, wrote the rules, ran them, kept notes, and wrote up reports.” — CloudSek
Inside the exposed server
The breakthrough came through a basic mistake. The group left one of its servers open to the internet, allowing CloudSek to recover data that showed, step by step, how the operation had been run. Among the files were the instructions the hackers fed to their AI agent, along with guidelines on how it should see itself, behave and communicate with its human handlers.

















