Eight AI Agents Launched an Attack: How Taiwan's Government Systems Were Breached

An unprecedented cyberattack using autonomous AI agents breached Taiwan's government systems. According to a Financial Times report based on research by the Israeli cybersecurity company Dream, attackers compromised at least 85 government employee accounts and exfiltrated over 2,500 sensitive records, using open-source AI systems named Hermes and OpenClaw. Researchers suspect a link to China due to the use of simplified Chinese in internal communications.

An unprecedented cyberattack using autonomous AI agents breached Taiwan's government systems. According to a Financial Times report based on research by the Israeli cybersecurity company Dream, the attackers created a hacking tool that behaved like a coordinated team of human hackers. The system compromised at least 85 government employee accounts and exfiltrated over 2,500 sensitive records, before expanding its activity to Taiwan's Nuclear Safety Agency and at least seven local energy companies. The hackers used open-source AI systems named Hermes and OpenClaw, bypassing the models' built-in safety mechanisms by defining the attack as an 'authorized penetration test'. The researchers emphasized that the use of simplified Chinese in the attack tool's internal communications likely indicates a connection to China, as the target itself uses traditional Chinese. Taiwan's Digital Ministry refused to officially comment on the breach details for confidentiality reasons, but noted that the integration of AI created a dual challenge for defense systems, as the attacks became fully automated.