'Near-autonomous' AI agents attack Taiwan's nuclear safety agency
Some say the world will end in fire, some say an agentic swarm
Chinese actors leveraged publicly available AI tools to infiltrate Taiwanese government systems, escalating the breach to include nuclear safety agencies, supply-chain vendors, and over seven energy companies in what security experts termed a near-autonomous assault. Between July 1 and July 4, AI agents compromised 85 government user accounts and obtained more than 2,500 personnel records, according to Israeli cybersecurity firm Dream.
The attack surface was mapped, revealing 85 government user accounts and extracting over 2,500 personnel records. Dream researchers, in their detailed report, discovered evidence of the operation in a 160 MB online archive containing 1,395 files. Taiwan was identified as the primary target, though the firm declined to disclose the specific government affected.
The AI-driven attack framework, based on open-source Hermes and OpenClaw agents, deployed up to eight sub-agents, each targeting distinct entities during 12 waves of attacks. The agents mapped the entire government ecosystem, extracting URLs, API endpoints, OAuth client IDs, and Keycloak configuration objects. One system exposed its entire user database without authentication, exposing thousands of employee records.
The attack framework utilized "learning cycles" to search vulnerability databases and GitHub repositories for exploit techniques. If the AI framework encountered errors, it self-corrected its actions. This near-autonomous attack underscores the growing sophistication of cyber threats and the potential for AI-driven autonomous attacks.
Written by urgent.news from The Register's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
This story
This is one outlet's version. Read the fullest account.
