51 AI Agents Used by CyberXero to Steal 628k Ukrainian Records via WordPress

A financially motivated threat actor known as CyberXero has exfiltrated more than 628,000 records tied to Ukrainian individuals, according to SOCRadar. The group, described as a Russian-speaking initial access broker, combined commodity offensive tools with an AI orchestration layer powered by Anthropic’s Claude Code. This operation targeted both global e-commerce platforms and specific Ukrainian energy entities, running two distinct pipelines simultaneously since at least July 2026.

The risk

The campaign’s scale was revealed when SOCRadar discovered an open directory at IP address 46.21.250[.]135 that lacked access controls. This misconfiguration exposed a live Linux workstation’s working directory, containing over 90,000 files. Researchers found AI session logs, scripts with plaintext tokens, exploitation toolkits, and already-stolen victim data in this exposed folder. The logs detailed every planned target, refused task, and workaround attempt, effectively providing a complete record of the attacker’s operations.

Who is exposed

The operation ran two parallel campaigns: a global, automated sweep against WordPress and e-commerce sites, and a curated, manual effort against Ukrainian critical infrastructure. The global pipeline used a custom tool called wp2shell to exploit known security flaws, including CVE-2026-63030, CVE-2026-60137, and CVE-2026-4815, an SQL injection vulnerability in the Support Board plugin. Analysis showed a single automated execution scanned 4,708 targets, gaining access to 429 WordPress administration panels and deploying 32 web shells within 61 seconds.

Cost and compliance

Confirmed compromises occurred in Ukraine, Poland, China, and the Netherlands, according to SOCRadar. The targeted Ukrainian organizations included energy providers and critical infrastructure entities, such as a heating provider in Kharkiv. The actor is assessed as an initial access broker, meaning they sell network access to other criminals rather than exploiting it directly for immediate financial gain. The campaign remains ongoing, with no indication that the group has ceased its activities.

What should you do?

Managers should immediately audit their web servers for exposed directories and ensure all directories require authentication. Developers must verify that WordPress sites and plugins, especially the Support Board plugin, are patched against known SQL injection vulnerabilities. Security teams should monitor for unusual outbound traffic that could indicate web shell deployment or data exfiltration.

Background: AI agents

AI agents are software systems that perceive their environment, make decisions, and execute actions to achieve specific goals. Unlike standard models, they maintain internal state and interact with external tools, creating unique security risks around tool execution, memory retention, and autonomous decision loops that require specialized controls.

Read the full guide: AI Agents Explained: Answers to Security and Architecture Questions

What to do and how to stay safe: CyberXero

• Scan your web applications for open directories that may expose sensitive files or configuration data without authentication.
• Update WordPress installations and plugins to versions that address CVE-2026-63030, CVE-2026-60137, and CVE-2026-4815.
• Monitor server logs for rapid, automated scanning patterns or the creation of unauthorized administrator accounts.
• Review access controls on any infrastructure hosting development tools or AI agent configurations to prevent accidental exposure.

Step-by-step guide: AI Governance Explained: Building Rules for Autonomous Systems

General security guidance from the Vector Tribune newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

What is CyberXero?

CyberXero is a Russian-speaking, financially motivated initial access broker that uses AI agents and commodity tools to compromise networks.

How many records were stolen?

More than 628,000 records tied to Ukrainian individuals were confirmed to be in the possession of the threat actor.

What tools did the attackers use?

The group used Cobalt Strike, Anthropic’s Claude Code for AI orchestration, and a custom tool called wp2shell for WordPress exploitation.