Cybersecurity firm Hunt.io says it uncovered a cyber-espionage campaign against Thailand’s Ministry of Finance that was largely orchestrated by an autonomous AI agent operating with no human in the loop. The finding came after researchers discovered a hacker-controlled server that had been left publicly accessible, exposing hundreds of files that documented the intrusion as it was still unfolding.

The exposed cache included malware, stolen credentials, attack scripts, AI agent logs, and evidence of access to multiple internal ministry systems. According to Hunt.io, much of the activity was driven by Hermes, an open-source AI agent released earlier this year by AI research company Nous Research. The attackers ran the tool in its unrestricted “YOLO mode,” which allows it to execute commands without waiting for human approval.

Researchers said the agent independently explored the ministry’s network, searched internal files, gathered system information, and hunted for privilege escalation opportunities. The attacker infrastructure also contained exploits for known software vulnerabilities, scripts custom-built for the ministry’s environment, active authentication cookies, and a previously undocumented backdoor family dubbed Hades, available in both Windows and Linux variants capable of remote command execution and file transfer.

Scripts recovered from the server referenced the ministry’s administrative web portals, email systems, and document management platforms, along with tools designed to test ministry email passwords. Hunt.io said it found no evidence that data had been exfiltrated; the activity instead appeared focused on reconnaissance, credential harvesting, and network mapping for potential follow-on operations.

  • Malicious activity traced back to at least mid-to-late June
  • ThaiCERT and Thailand’s National Cyber Security Agency notified July 15
  • Multiple indicators pointed to Chinese-speaking operators, though no group has been attributed

The Finance Ministry has not publicly acknowledged the incident. Following the disclosure, Thai cybersecurity officials said they would strengthen national defenses against AI-agent-driven attacks, with deputy secretary-general Theerawut Wittayakorn stating that Thailand must manage AI-related risks while still benefiting from the technology.

The incident follows a separate disclosure this month in which Hugging Face said an autonomous AI agent, later confirmed to belong to OpenAI, breached its systems using two previously unknown vulnerabilities and stolen credentials. Together, the cases mark an escalation in adversaries’ use of unsupervised AI agents to conduct real-world intrusions.