AI Agents Breach Taiwanese Government and Energy Systems

The Ministry of Digital Affairs of Taiwan confirmed that government agencies were targeted by AI-assisted cyberattacks from an overseas source in July 2026. The breach utilized a hybrid approach combining manual operations with autonomous AI agent frameworks, specifically Hermes and OpenClaw, which operated largely without human direction to map 21 government systems and compromise at least 85 accounts.

An investigation by the Israeli cybersecurity firm Dream revealed that the agents stole over 2,500 personnel records from the justice ministry, extracted passwords from officials, and scanned a nuclear safety agency for vulnerabilities. The attack also expanded into seven energy companies. While the Taiwanese government did not explicitly name China, Dream identified internal messages in simplified Chinese, and the incident follows reports from the National Security Bureau of a 6% increase in Chinese cyberattacks on key infrastructure in 2025.

The National Institute of Cyber Security issued warning alerts starting July 20, and the Ministry of Digital Affairs stated that affected units have since completed their handling of the incident. In response, the government has implemented new protective guidelines and strengthened system monitoring. The event has sparked industry-wide concern, with cyber insurance underwriters questioning if current policies cover unsupervised AI attacks.