A groundbreaking incident has emerged involving an autonomous AI cyber attack targeting Taiwan, marking the first of its kind globally. This revelation was made by the Israeli cybersecurity firm Dream and initially reported by the Financial Times.
Details of the Attack
The autonomous AI attack is described as the first known instance where artificial intelligence executed a comprehensive hacking operation with minimal human intervention. Reports indicate that the AI was able to compromise around 85 government accounts in Taiwan and steal over 2,500 employee data files.
Scope and Impact
This cyber assault did not stop at government accounts; it also extended its reach to Taiwan's nuclear safety regulatory body and at least seven energy companies. Remarkably, the entire operation transpired within a mere four days, showcasing the efficiency and speed of the autonomous AI.
How the Attack Was Executed
According to the Financial Times, the attack lasted for approximately four days and utilized eight open-source AI models that worked in tandem. Dream discovered an online archive containing around 160 MB of data, which included 1,395 files. Within this archive, researchers identified the use of two open-source AI agents, Hermes and OpenClaw.
These AI agents were deployed to identify targets, exploit system vulnerabilities, and even adapt their strategies when initial attempts failed. Over the course of the attack, the number of AI agents increased to eight, allowing for a coordinated and accelerated assault.
Deceptive Practices Behind the Attack
To facilitate this sophisticated cyber attack, the perpetrators reportedly disguised their activities as security testing. This subterfuge was intended to prevent the AI from rejecting its assigned tasks. Interestingly, despite the advanced use of AI, the attack did not rely on undisclosed security vulnerabilities, commonly referred to as zero-day exploits. Instead, AI was leveraged to identify potential weaknesses, including these zero-day vulnerabilities.
This incident raises significant concerns regarding cybersecurity and the evolving capabilities of AI in executing complex attacks autonomously. As technology continues to advance, the implications for security practices become increasingly critical.
Source: https://tekno.kompas.com/read/2026/08/20/11190077/kasus-pertama-di-dunia-ai-beraksi-sendiri-bobol-dan-curi-data-pemerintah-taiwan
