China-Linked Hackers Conduct First Fully Autonomous AI Cyberattack on Taiwan Government

In a watershed moment for cybersecurity, researchers at Israeli AI and cyberdefense firm Dream documented what appears to be the first fully autonomous, end-to-end AI-powered hacking campaign targeting a government entity. Operating over four days in early July 2026, suspected China-linked attackers deployed up to eight autonomous AI agents simultaneously using publicly available frameworks known as Hermes and OpenClaw to systematically breach Taiwanese government infrastructure. The attack demonstrated unprecedented coordination and autonomy. The agents mapped 21 government systems, researched vulnerabilities, and adapted tactics whenever blocked—each agent working different attack vectors as though coordinated by human operators. When one infiltration route failed, another agent independently searched the internet for fresh intelligence and devised alternative approaches, allowing the operation to advance without constant human direction. The framework bypassed safety mechanisms built into the underlying AI models by framing the intrusion as an authorized penetration test. The scope of the breach expanded far beyond initial targets. By the time researchers discovered the operation, the agents had compromised at least 85 government user accounts and exfiltrated more than 2,500 personnel records, seven SSO client secrets, six internal database credentials spanning multiple systems, and a complete JSON export of all department system users. The operation then pivoted outward, scanning Taiwan's nuclear safety agency, at least seven energy companies, government IT supply-chain vendors, and government email systems simultaneously for misconfigurations and exploitable vulnerabilities. Taiwan's Ministry of Digital Affairs officially confirmed the attack on August 13, 2026, emphasizing that AI agents could now rapidly combine multiple attack techniques and use secondary systems as stepping stones, making attacks faster, cheaper, and dramatically more scalable. The discovery raised urgent questions about global preparedness for AI-driven warfare, with Taiwanese officials noting that defensive strategies, legal frameworks, technical capabilities, and policies remain insufficient in most countries. The incident coincided with Taiwan already facing sustained pressure from China—the island's National Security Bureau reported an average of 2.6 million cyberattacks from China daily in 2025, up 6% year-over-year. If a significant fraction of that volume begins operating with this level of autonomy, defenders face a fundamentally transformed threat landscape where human-scaled incident response may become inadequate.