Skip to content
Submit incident
AII-1000-30487
Under investigationA regulator, court or police has opened a formal inquiry and reached no finding yet.

AI Agent Used in Post-Compromise Activity Against Thailand's Ministry of Finance

An attacker used Hermes, an open-source AI agent running in YOLO mode, to autonomously conduct post-compromise reconnaissance and attack preparation on Thailand's Ministry of Finance network. Security researchers discovered the attacker's exposed logs and tooling after Thai authorities were notified in July 2026, though no data exfiltration was confirmed.

June 2026 Β· ThaiCERT Β· Security and cyberattacks
June 2026
AII-1000-304872 sources ↓
Quick facts
Category
Security and cyberattacks
Country
Thailand
Source
thehackernews.com
Date
June 2026
Drafted by
Raidu Incident Detection

What happened

An attacker installed Hermes, an open-source AI agent, on a rented server and enabled its YOLO mode, which allows execution of commands without human approval. The operator directed the agent at Thailand's Ministry of Finance network, which the attacker had already infiltrated through unknown means.

The agent autonomously explored the ministry's network, checked hosts for root access opportunities, searched file systems, and accessed a folder of staff personnel records dating back to 2012. The operator left the agent's logs and 585 files containing 470 MB of attack tooling on a publicly accessible web server, which security firm Hunt.io and researcher Bob Diachenko discovered.

Reported impact

The exposed infrastructure contained malware, stolen credentials, attack scripts, AI agent logs, exploits targeting known software vulnerabilities, active authentication cookies, and a custom backdoor malware family called Hades capable of remote command execution and file transfer. Hunt.io found evidence the attackers had compromised multiple ministry systems, though no data exfiltration was detected.

Index analysis

The Hermes agent's YOLO mode is a documented feature, not a flaw in the tool. The operator conducted human-directed reconnaissance, building password lists from the ministry's own department abbreviations and embedding ministry-specific intranet paths in attack scripts. The agent then autonomously executed these prepared commands within the compromised network.

Governance controls

Thailand's national CERT and National Cyber Security Agency were notified on July 15. The Ministry of Finance had not publicly acknowledged the incident at the time of reporting.

Classification

Category
Security and cyberattacks
Industry
Government, security and defence
Country
Thailand
Harm status
Actual harm
Intent
Intentional
Timing
Post-deployment
Harm types
Public interest
Harmed
Government
Failure modes
Unsafe autonomous behaviour
AI systems
Hermes

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

thehackernews.com
AI Agent Used in Post-Compromise Activity Against Thailand's Ministry of Finance
therecord.media
AI Agent Used in Post-Compromise Activity Against Thailand's Ministry of Finance