AI agent without confirmations: how Hermes was used for post-exploitation in Thailand’s Ministry of Finance network

Colleagues, I’d like to highlight a cybersecurity case: the Hermes AI agent was run in no-confirmation mode and deployed into Thailand’s Ministry of Finance network.

According to researchers, the agent carried out repetitive tasks on its own:
— scanned hosts for routes to root access;
— checked file systems and permissions;
— browsed directories containing employee data.

Importantly, this was not a Hermes product vulnerability. The risk arose from disabling oversight and using standard tools to automate post-exploitation.

Why it matters: such scenarios show how AI can accelerate an already underway attack and reduce operator workload.

Do you think teams are ready for this attack model?

#cybersecurity #AI #ThreatIntel #Hacking