AI Signal 441
Near-autonomous AI agents compromise Taiwan government systems and nuclear safety agency
↗
Suspected Chinese operatives used AI agents built on open source frameworks to autonomously compromise 85 Taiwanese government accounts, exfiltrate over 2,500 personnel records, and expand attacks to a nuclear safety agency and energy companies.
This is a documented case of AI agents conducting near-autonomous cyber operations against critical infrastructure, including self-correction and autonomous learning cycles that searched vulnerability databases for exploitation techniques. The attack demonstrates that publicly available AI tools can now automate reconnaissance, credential harvesting, CAPTCHA bypass, and lateral movement across government ecosystems with minimal human direction.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
AI agents built on open source Hermes and OpenClaw frameworks deployed eight sub-agents across 12 attack waves, compromising 85 government accounts and exfiltrating over 2,500 personnel records.
The agents autonomously discovered unauthenticated API endpoints, solved CAPTCHAs with 100 percent accuracy, and cracked passwords through predictable pattern testing.
The attack pivoted from government systems to supply chain vendors, a nuclear safety agency, and seven or more energy companies, scanning them in parallel for misconfigurations and vulnerabilities.
THE CLUSTER