CyberXero uses AI agents to steal 628,000 Ukrainian records

Cybersecurity researchers have identified a threat actor known as CyberXero that utilizes an AI orchestration layer to conduct large-scale cyberattacks. The group, described as a Russian-speaking, financially motivated initial access broker, uses 51 AI agents—including Claude Code from Anthropic—to manage its operations.

The actor operates two distinct pipelines: a global, automated campaign targeting WordPress sites and e-commerce platforms through known vulnerabilities, and a manual, selective campaign targeting Ukrainian energy and critical infrastructure. The latter has resulted in the confirmed theft of over 628,000 Ukrainian records, including sensitive data from residents in Kharkiv.

Researchers from SOCRadar discovered the operation after identifying a misconfigured server with an open directory. This exposure revealed more than 90,000 files, including AI session logs, scripts containing plaintext tokens, and stolen victim data. In addition to the attacks in Ukraine, breaches have been confirmed in Poland, China, and the Netherlands.