Suspected Chinese cyber operators reportedly used publicly available AI tools in a “near-autonomous” campaign against Taiwanese government systems, later targeting the nuclear safety agency, suppliers and energy companies. The case matters because it suggests AI agents can coordinate reconnaissance, credential attacks and data theft at speed, potentially widening the impact of a single intrusion.

Dream, an Israeli cybersecurity firm, said the operation ran from 1 to 4 July in 12 attack waves using up to eight sub-agents built on the open-source Hermes and OpenClaw tools. It allegedly compromised 85 government accounts, extracted more than 2,564 personnel records and obtained credentials, client secrets and internal network information; Taiwan was identified by a source familiar with the attack, though Dream did not publicly name the government or attribute the operation conclusively to China.