Taiwan targeted in autonomous AI agent attack

Hackers used AI agents to run an automated attack on a government target understood to be Taiwan, according to Israeli AI and cybersecurity company Dream. Researchers uncovered the July campaign after finding a 160MB online file containing more than 1,000 files that showed how Hermes and OpenClaw were used to conduct two simultaneous, autonomous attack campaigns over four days.

The AI models behind the operation were not identified, but recovered data indicated their safeguards were bypassed by framing the activity as an authorised security test. Internal communications were in Simplified Chinese, which Dream said pointed to a high probability of links to the People’s Republic of China, while data taken from the target government was in Traditional Chinese.

Dream said the agents compromised at least 85 government user accounts and stole more than 2,500 personnel records before targeting Taiwan’s nuclear safety agency and at least seven energy companies. The company described the incident as the first it had identified involving autonomous agents attacking a government.