A recent cyber attack on Taiwan's nuclear safety agency has raised concerns about the increasing use of near-autonomous AI agents in cyber warfare. The attack, which occurred over four days in July, was carried out by suspected Chinese cyber operatives who used publicly available AI tools to compromise Taiwanese government systems, including supply-chain vendors and energy companies.
Key Insights
The attack framework used open source Hermes and OpenClaw AI agents, which were able to deploy up to eight sub-agents, highlighting the complexity and sophistication of the attack. The use of near-autonomous AI agents allowed the attackers to compromise 85 government user accounts and extract over 2,500 personnel records, demonstrating the potential of AI-powered cyber attacks to cause significant harm.
