INSPECTING ARCHIVED INTELLIGENCE (OUTDATED VERSION).

Chinese Hackers Utilize AI Agents in Multi-Country Cyber Campaign

| 2026-09-03 17:26 MEDIUM MEDIUM AI-ENABLED ATTACK STATE-SPONSORED & ESPIONAGE
JSON
Executive Summary
AI-generated
A second, separate China-linked campaign was documented by threat intelligence firm [IOC HIDDEN • LOGIN REQUIRED] on September 4, 2026, where commercial AI models were wired into live cyberespionage operations targeting Taiwan's Kuomintang Party archives and government systems in mainland China. The targeted sectors included education and industrial hosts in Vietnam. This attack worked by using a framework called SecFlow that utilized different AI models such as Claude, Qwen, and DeepSeek to automate traditional hacking tasks like scanning for vulnerabilities, testing stolen credentials, deploying webshells, collecting data and evidence, and generating reports. The current status indicates the most damaging breach hit a Fengtai District government Office Automation environment in China, resulting in command execution, collected sensitive information, and deployed multiple Windows implants.
Technical Mitigations AI-generated
• <a href="/auth/login?next=/detail/4GYZcKAB-1kL6CVYv7NC" class="ioc-censored-pill text-decoration-none" title="Protected IoC: Sign in to view" data-bs-toggle="tooltip"><span class="badge bg-black text-warning border border-warning border-opacity-75 font-monospace ioc-lock-tag align-middle"><i class="bi bi-lock-fill me-1"></i>[IOC HIDDEN &bull; LOGIN REQUIRED]</span></a> uncovered a Chinese-speaking campaign using AI agents to automate cyberattacks against Asian government, education and industrial targets. • The framework behind the campaign, called SecFlow by the operators, could use different AI models, including Claude, Qwen, and DeepSeek. • Researchers reconstructed the entire orchestration system from five exposed open directories left publicly accessible by the operators.
Technical Observables Login Required
Indicators of Compromise (IoCs) Protected

Raw threat telemetry, malicious IP addresses, file hashes, and direct VirusTotal correlation are restricted to authenticated users.

Sign In to View 25 IoCs Create Free Account
ni•••••.com
ws•••••.run
hu•••••.io
ex•••••.aspx
hxxp://••••••••••••••••••••
9ef858••••••••••••••••••••••••••••••••••••••••••••••••••••••••••
4ecbda••••••••••••••••••••••••••••••••••••••••••••••••••••••••••
79cc58••••••••••••••••••••••••••••••••••••••••••••••••••••••••••
135b33••••••••••••••••••••••••••••••••••••••••••••••••••••••••••
Intelligence Metadata
Actors / Malware / CVEs / Campaigns
Campaign Pierluigi PaganiniCampaign Pierluigi Paganini
Target & Sectors
ASEAN ASEAN FIVE_EYES FIVE_EYES educationeducation governmentgovernment
Incident Timeline
‎2026/09/03
A Chinese-speaking operator used AI agents to automate cyberattacks against Asian government, education and industrial targets.
organisation SecFlow
A Chinese-speaking threat operator deployed an AI orchestration framework called SecFlow to conduct intrusions across Taiwan, Indonesia, China, Vietnam, and Afghanistan.
organisation Kuomintang Party History Archives
Targets included Taiwan's Kuomintang Party History Archives, Indonesia's Ministry of Foreign Affairs, government and education systems in mainland China, and industrial hosts in Da Nang, Vietnam.
organisation Ministry of Foreign Affairs
Targets included Taiwan's Kuomintang Party History Archives, Indonesia's Ministry of Foreign Affairs, government and education systems in mainland China, and industrial hosts in Da Nang, Vietnam.
organisation SecBox
The most significant compromise affected a Fengtai District government environment, achieving command execution, credential theft, and deployment of SecBox implants.
infrastructure Windows
“The most extensive compromise hit a Fengtai District government environment, where the operator achieved command execution, collected LSASS and registry hives, accessed government and health records, and deployed multiple Windows implants.
“ Operators achieved Windows command execution through internet-facing web applications, then used ASPX web shells not just to run commands but as a persistent operational backbone: querying internal databases, pulling an LSASS memory dump apart…
Another server-side page, extract.aspx, scanned sections of the memory dump for Windows password-hash material.”
organisation Claude
The campaign utilized commercial AI models (Claude, Qwen, DeepSeek) as operational components for reconnaissance, exploitation, and data collection.
organisation Shellshock, Log4Shell
The operator exploited eight CVEs including Shellshock, Log4Shell, and Spring4Shell, deploying steganographic GLUTTON webshells concealed in PNG images.
organisation PNG
The operator exploited eight CVEs including Shellshock, Log4Shell, and Spring4Shell, deploying steganographic GLUTTON webshells concealed in PNG images.
data_breach 822 OA account records
From there they pulled 822 OA user account records, created a new privileged account of their own for backup access, and walked away with 949 attachments totaling 1.28GB, including a chronic-disease report containing real patient health information.
data_breach 1.28 GB
From there they pulled 822 OA user account records, created a new privileged account of their own for backup access, and walked away with 949 attachments totaling 1.28GB, including a chronic-disease report containing real patient health information.
data_breach 104 complete chatbot conversations
It exposed 23 AI agent configurations, 14 API secret fields containing credentials, and 104 complete chatbot conversations.
Tactical Metrics
Metrics
infrastructure
‎Windows
Affected Product
“The most extensive compromise hit a Fengtai District government environment, where the operator achieved command execution, collected LSASS and registry hives, accessed government and health records, and deployed multiple Windows implants.
“ Operators achieved Windows command execution through internet-facing web applications, then used ASPX web shells not just to run commands but as a persistent operational backbone: querying internal databases, pulling an LSASS memory dump apart…
Another server-side page, extract.aspx, scanned sections of the memory dump for Windows password-hash material.”
Metrics
data_breach
822
Oa Account Records
From there they pulled 822 OA user account records, created a new privileged account of their own for backup access, and walked away with 949 attachments totaling 1.28GB, including a chronic-disease report containing real patient health information.
Metrics
data_breach
1
Gb
From there they pulled 822 OA user account records, created a new privileged account of their own for backup access, and walked away with 949 attachments totaling 1.28GB, including a chronic-disease report containing real patient health information.
Metrics
data_breach
104
Complete Chatbot Conversations
It exposed 23 AI agent configurations, 14 API secret fields containing credentials, and 104 complete chatbot conversations.
Intelligence Sources
Security Affairs 2026-09-04
AlienVault OTX 2026-09-03
AlienVault OTX 2026-09-04