UAT-10147 is a financially motivated Chinese-speaking cybercrime group identified by Cisco Talos in early 2026. The group targets a broad range of vulnerable Windows and Linux web servers worldwide, leveraging publicly disclosed vulnerabilities for initial access. They integrate agentic AI systems into their post-compromise operations, automating exploitation, reconnaissance, payload generation, validation, and persistence workflows. The actor employs open-source offensive frameworks such as Metasploit, ysoserial, PentestGPT, and DeepAudit, alongside multiple privilege escalation exploits, to automate intrusion and establish persistence. Their AI-driven approach includes iterative exploit refinement, adaptive troubleshooting, exploit validation, and operational documentation generation, representing a transition from AI-assisted scripting to semi-autonomous offensive orchestration. The campaign targets sectors including government, education, media, technology, and gaming, with affected servers identified in Brazil, Bolivia, China, Canada, and Vietnam. The actor’s operational security failure revealed a download server with an open directory, providing insight into their infrastructure and target list of approximately 170,000 URLs. Post-compromise, the group deploys malware for SEO fraud and data theft or installs web shells to manually deploy additional backdoors such as BadIIS malware.

The threat actor achieves remote code execution on vulnerable Windows and Linux web servers, enabling deployment of malware for SEO fraud and data theft. The integration of AI-driven tooling enhances the scale, efficiency, and sophistication of attacks, reducing the expertise required for complex post-compromise operations. This increases the risk of widespread compromise of high-value internet-exposed servers across multiple sectors globally. The actor’s ability to automate exploit refinement, validation, and persistence workflows may lead to more persistent and harder-to-detect intrusions. The campaign’s impact includes unauthorized data access, potential data exfiltration, and abuse of compromised servers for fraudulent activities.

Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. Organizations should prioritize applying security updates for publicly disclosed vulnerabilities affecting their Windows and Linux web servers. Monitoring for indicators of compromise related to UAT-10147, such as unusual remote code execution attempts, deployment of web shells, or presence of BadIIS malware, is recommended. Restricting internet exposure of critical web servers and implementing strong access controls can reduce attack surface. Given the actor’s use of AI-driven automation, defenders should consider enhanced detection capabilities for automated exploitation patterns. Review and harden server configurations to prevent unauthorized code execution and persistence mechanisms.