Eight AI agents launched an attack: how government systems in Taiwan were breached

Hackers suspected of being linked to the Chinese government recently used free artificial intelligence tools to breach government websites in Taiwan, a disclosure that demonstrates how AI technology is fundamentally changing the face of modern cyber warfare.

According to a report by the Financial Times, based on research by the Israeli cyber company Dream, the attackers created an autonomous hacking tool that behaved effectively like a coordinated team of human hackers.

During four days in early July, the system deployed up to eight autonomous AI agents simultaneously. These agents mapped 21 different government systems, identified security vulnerabilities, and independently changed operational tactics when encountering blocks.

As part of the attack, the system breached at least 85 government employee accounts and extracted more than 2,500 sensitive records before expanding its activity to Taiwan's Nuclear Safety Agency and at least seven local energy companies.

Dream, founded by cyber entrepreneur Shalev Hulio and former Austrian Chancellor Sebastian Kurz, identified the breach after locating a network archive containing 1,395 files.

The research shows that the hackers used open-source AI systems named Hermes and OpenClaw. The attackers bypassed the models' built-in safety mechanisms by defining the system as a "sanctioned penetration test," a justification that caused the artificial intelligence to cooperate with the attack without restrictions.

Amir Becker, the company's Chief Strategy Officer and a former senior official in Unit 8200, noted:

"I have never seen an end-to-end autonomous attack against a government target of this magnitude."

The company refused to officially link the attack to a specific group, but researchers emphasized that the use of simplified Chinese in the internal communication of the attack tools points with high probability to a connection to China, as the target itself uses traditional Chinese.

Taiwan's Ministry of Digital Affairs declined to comment on the details for confidentiality reasons, but noted that the integration of AI has created a double challenge for defense systems, as the attacks have become fully automated.

The island currently faces an average of about 2.6 million Chinese cyberattacks daily, and this disclosure makes it clear that the era of independent artificial intelligence attacks is only in its infancy.