Chinese-suspected hackers deployed an autonomous artificial intelligence system over four days in July to execute a cyberattack targeting Taiwanese government infrastructure. Discovered by Israeli cybersecurity firm Dream, the open-source AI agents autonomously mapped 21 government systems, compromised 85 user accounts, and exfiltrated 2,500 personnel records in the first known fully autonomous cyber operation against state agencies.
This operational breakthrough relied on multi-agent frameworks, including OpenClaw, to automate reconnaissance, execute credential attacks, and dynamically map subsequent attack vectors without direct human intervention. Taiwan’s Ministry of Digital Affairs confirmed that the overseas cyber intrusion combined conventional hacking tactics with autonomous software agents, exposing critical structural vulnerabilities in public sector digital defenses. The attack signals a key technical evolution in asymmetric warfare, demonstrating how state-aligned threat actors leverage accessible machine-learning frameworks to drastically accelerate the overall speed, scale, and operational efficiency of offensive cyber campaigns worldwide.