Suspected China-linked hackers used open-source artificial intelligence agents to breach Taiwanese government networks over four days in early July, according to a Financial Times report citing research from Israeli cybersecurity firm Dream. Researchers said the intrusion mapped 21 government systems, compromised at least 85 user accounts and extracted more than 2,500 personnel records before expanding to Taiwan’s Nuclear Safety Commission and at least seven energy companies. Dream stopped short of publicly naming Taiwan, citing company policy, saying only that it had notified a government in the Asia-Pacific region. A person familiar with the matter confirmed the target was Taiwan.
Dream’s investigation, which recovered a 160-megabyte online archive containing 1,395 files, found the operation ran on two open-source agentic frameworks: Hermes, built by Nous Research, and OpenClaw, created by developer Peter Steinberger. Researchers said as many as eight AI agents operated in parallel after being deployed by human operators, then independently conducted reconnaissance, identified vulnerabilities and adapted when attack paths were blocked. Dream said the agents bypassed built-in safety guardrails by framing the intrusion internally as an authorized penetration test.
Researchers said they could not determine which underlying AI model powered the agents. They pointed to Simplified Chinese in the attackers’ internal communications and Traditional Chinese formatting in the compromised government data as indicators consistent with a China-based operator, though Dream has not formally attributed the campaign to a specific hacking group or to the Chinese government.
Taiwan’s Ministry of Digital Affairs declined to comment on the specific incident, citing confidentiality, but said AI-enabled attacks now pose a “dual challenge” because autonomous agents can both automate cyber intrusions and become attack targets themselves. Beijing had not issued a public response as of Tuesday. Taiwan’s National Security Bureau reported in January that the island faced an average of 2.6 million Chinese cyberattacks per day in 2025, a 6% increase from the previous year. Separately, AI company Anthropic disclosed in November that it had detected China-linked actors attempting to misuse its Claude models against around 30 organizations, though the company said those efforts achieved only limited success.
