Taiwan's Ministry of Digital Affairs says the attack on government agencies began on 20 July, and Israeli firm Dream, which detected the intrusion, calls it a first-of-a-kind breach. Attackers used open-source AI agents — OpenClaw and Hermes — to build what Dream calls an autonomous hacking tool, one that behaved like a coordinated cyber team.
It compromised at least 85 government accounts, extracted more than 2,500 personnel records, then expanded to Taiwan's nuclear safety agency and at least seven energy companies. Not everyone is sold on the autonomy: Semgrep's Cris Thomas points out that someone still picked the target and set the objective — "it's not totally 100% autonomous."
While Taiwanese officials stopped short of blaming China, Dream said the use of Simplified Chinese in internal communications meant a high probability the operator was connected to China. China's Taiwan Affairs Office did not respond to a request for comment.
MDA says the attack combined manual operations with AI agents, that sources, methods and scope have been fully investigated, and that affected units have completed their handling. New protective guidelines and strengthened monitoring are now in place.
Source: The Guardian