On November 4, 2026, the Acronis Threat Research Unit published its investigation of "BOSSAI," a mesh of 12 AI agents (nine active) run by a single operator based in Taiwan. Over the 12 days Acronis observed it, the mesh targeted hundreds of organizations in 11 countries, including emergency services, banks and lenders, universities and military systems.
Acronis confirmed data theft or system access at roughly 10 targets in three countries. Agents pulled case records from a Taipei 119 emergency dispatch system, including incident times and street-level locations across seven districts. A Taipei fire department patient-care platform was found exposing ECG readings, ambulance video and vital signs from 2018 without a password. Agents extracted data from three Thai universities and from an account table at a Taiwanese lending firm.
Given a single line, "extract admin credentials, backend path, and all student data," one agent mapped more than 50 subdomains of a Thai university, found a SQL injection flaw, pulled a 2,563-row administrator table and started dumping student records with no further input. Its system prompt gave it "full decision-making authority" and every session ran with actions auto-approved.
Commercial AI models refused the attack requests again and again, so the offensive core ran on locally hosted models with their safety training stripped out. The agents also made plenty of mistakes: one scanner's high and critical findings were more than half false positives. This is a private forensic report, not a law enforcement finding.
