hackingincidents.ai
Incidents289Sources240Dataset2026-11-17Last sweep2026-11-17
Ledger / Jul 2026

Open-source Hermes and OpenClaw agents used against the Taiwanese government

Confirmed

Dream "uncovered the complete operational workspace" (1,395 files) of a Hermes and OpenClaw framework that "achieved confirmed, real-world compromises against state infrastructure" on 1-4 July 2026. That is direct observation. CyberScoop quotes Dream's post verbatim ("The attacker didn't stop at primary targets ...") and credits the Financial Times with naming Taiwan as the target. Dream itself names no country: the grade rests on Dream's evidence, and the Taiwan link on the Financial Times.

Model
Not namedHermes, OpenClaw
Agent harness
Hermes, OpenClawopen-source harness
Attributed country
Not attributedthe source names no country
Sectors hit
governmentTaiwanese government agencies (85 credentials, 2,564 personnel records)

Attackers used the open-source Hermes and OpenClaw agent frameworks against government agencies, and Dream recovered the framework's operational workspace. The Taiwan attribution comes from the Financial Times via CyberScoop; Dream's own report names no country. Taiwan's Ministry of Digital Affairs described a hybrid of manual and AI agent-assisted attacks.

“The hackers used two popular open-source AI frameworks, Hermes and OpenClaw, to set up the Taiwan operation.”CyberScoop

Sources

Date
Jul 2026
Actor
Suspected China-linked hackers (Taiwan did not attribute) (unknown), as attributed by CyberScoop
Nexus
China (suspected; not attributed by Taiwan)
Targets
Taiwanese government agencies (85 credentials, 2,564 personnel records)
What the AI did
Agentic intrusion
Model as named
Hermes, OpenClaw
Model basis
A researcher names it
Autonomy
Semi-autonomous
Outcome
Taiwanese government agencies were attacked from 1 to 4 July 2026.