The world’s first autonomous AI agent cyberattack, likely involving operators liked to China and discovered by researchers at the Israeli AI company Dream, targeted Taiwan. That’s according to an article by the UK’s Financial Times dated August 12, citing a source close to the matter.

The article noted that attackers deployed up to eight AI agents to map government systems and research vulnerabilities. The tools were able to change tactics when blocked. The attack affected 85 government user accounts and extracted more than 2,500 personnel records before expanding to Taiwan’s nuclear safety agency and energy companies.

Dream declined to confirm which government had been targeted, and did not attribute the attack to any specific group. Researchers noted, however, that internal communications associated with the hacking operation used simplified Chinese characters, suggesting the operators were connected to China.

Taiwan’s Digital Ministry issued a statement on Thursday, August 13, saying that cybersecurity monitoring units detected attacks targeting government agencies in July. The ministry’s Administration for Cyber Security began issuing advisories on the matter starting July 20 and activated its investigation and response mechanisms. 

The ministry said that this wave of attacks showed clear signs of overseas origin and a pattern in which attackers used AI agents like OpenClaw to carry out the attacks. It said that AI agents can rapidly chain together multiple attack techniques and use secondary systems like backup or test environments as stepping stones for intrusion. This made the attacks fast, cheap, and scalable.

The ministry stated that in addition to the investigation and remediation of this incident, it will continue to strengthen the government’s overall cybersecurity defense capabilities based on the identified patterns. It will also continue to monitor other potential attack vectors to prevent similar incidents from occurring in the future.