Israeli cybersecurity firm Dream documented what it says is the first publicly known near-autonomous AI-powered cyberattack against a government, targeting Taiwan over four days in July. The attackers used open-source AI agent frameworks, including Hermes and OpenClaw, to compromise 85 accounts, extract more than 2,500 personnel records, and expand the intrusion to Taiwan’s nuclear safety agency and at least seven energy companies. The AI framework conducted parallel vulnerability scanning and self-corrected its own errors with limited human direction, though researchers noted the operation still required significant human setup and refinement.
