- Published on
CyberXero Uses Claude Code in Campaign That Stole Over 628,000 Ukrainian Records
- Authors
- Name
- DeepHorus
Cybersecurity researchers have disclosed an AI-augmented campaign attributed to CyberXero, a Russian-speaking, financially motivated broker. Active since at least July 2026, the operation targets WordPress sites, e-commerce platforms, and Ukrainian critical infrastructure. SOCRadar published the findings last week, saying the campaign remains ongoing.
The operation runs two pipelines at once, according to SOCRadar. A global automated campaign exploits WordPress and e-commerce flaws for access and payment data. A curated manual campaign conducts deep reconnaissance and exploitation against Ukrainian energy and utilities, including a Kharkiv heating provider.
The global campaign exploits wp2shell, tracked as CVE-2026-63030 and CVE-2026-60137, and CVE-2026-4815 in the Support Board plugin. Researchers confirmed exfiltration from four Ukrainian organizations, yielding more than 628,000 records tied to Ukrainian individuals. Other compromises occurred in Poland, China, and the Netherlands.
SOCRadar discovered the operation after finding an exposed open directory at 46.21.250[.]135. It leaked AI session logs, plaintext tokens, exploitation toolkits, reconnaissance output, agent configurations, and victim data. A single automated execution scanned 4,708 targets, confirmed access to 429 WordPress admin panels, and deployed 32 web shells in 61 seconds.
CyberXero deployed AI in two stages. It used Claude Code with 51 specialized agents for scanning, exploit chaining, reconnaissance, and post-exploitation. It also connected PentAGI to Cobalt Strike via an AI provider API for payload generation and pentest execution. To bypass refusals, the actor used context engineering and session resets, claiming an authorized pentest.
SOCRadar noted that safety controls act on session context, not model weights, so refusals do not carry across sessions. Logs still recorded genuine refusals: the model declined to install a backdoor, disable a firewall, move laterally, and deploy a web shell. The campaign also breached two Chinese entities via exposed Redis servers and remains active, urging operators to patch and monitor.
