Researchers observe first ‘near-autonomous’ AI attack on government target in Taiwan

Researchers at Israeli cyber firm Dream reported a suspected Chinese APT used open-source AI frameworks (Hermes and OpenClaw) to run a near‑autonomous multi‑agent intrusion against Taiwanese government infrastructure and related vendors, exfiltrating more than 2,500 personnel records and scanning multiple critical targets (government IT suppliers, a nuclear safety agency, email systems, and energy companies) while adapting its methods mid‑operation.