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

BOSSAI operator coordinates AI agents to breach emergency-service and university systems

Confirmed

Acronis examined the operator's agent conversations, commands, tools and victim data, directly linking AI agents to real intrusions.

Model
DeepSeek, GLM, QwenDeepSeek · open-weight; Zhipu AI · open-weight; Alibaba · open-weight
Agent harness
Hermes, Cursor, Trae, CyberStrike, Claude Code, Codex, OpenClawcustom framework
Attributed country
Taiwanas attributed by Acronis
Sectors hit
government, education, healthcareOrganizations across 11 countries; confirmed access or data theft at approximately 10…

One operator manually coordinated 12 specialized agents, relying largely on locally hosted models with safeguards removed. Agents scanned systems and attempted exploitation, while the operator transferred findings between interfaces. The date records the disclosure month; the operation's start month is unspecified.

“So the attacks ran mainly on "abliterated" models: freely downloadable (open-weight) versions of DeepSeek, GLM and Qwen with their safety training removed.”Acronis

Sources

Date
Nov 2026
Disclosed
8 Nov 2026
Actor
Unnamed Taiwan-based BOSSAI operator (unknown), as attributed by Acronis
Nexus
Taiwan
Targets
Organizations across 11 countries; confirmed access or data theft at approximately 10 targets in Taiwan, Thailand and Mexico, including universities and emergency services
What the AI did
Recon, Vuln research, Exploit dev, Agentic intrusion
Model as named
DeepSeek, GLM, Qwen
Model basis
A researcher names it
Autonomy
Semi-autonomous
Outcome
Confirmed system access and data theft, including medical information and personally identifiable information; many other targets were only scanned.