CyberXero AI Campaign Exfiltrates 628k Ukrainian Records via Exposed Server

SOCRadar researchers identified a campaign by CyberXero that exfiltrated more than 628,000 records tied to Ukrainian individuals. The threat actor, a Russian-speaking initial access broker active since July 2026, used AI orchestration alongside traditional tools to breach targets in Ukraine, Poland, China, and the Netherlands, according to the report published last week.

How did researchers discover the campaign?

SOCRadar found an open directory at IP address 46.21.250[.]135 that lacked access controls. This misconfiguration exposed the contents of a Linux workstation, including more than 90,000 files. The leak included AI session logs, scripts containing plaintext tokens, exploitation toolkits, reconnaissance data, and already-stolen victim information, providing a complete view of the operation.

What tools and tactics were used?

The operation runs two pipelines simultaneously: a global automated campaign and a manual campaign targeting Ukrainian energy infrastructure. The global effort uses wp2shell to exploit flaws in the WordPress REST API, specifically CVE-2026-63030, CVE-2026-60137, and CVE-2026-4815 in the Support Board plugin. Analysis showed one automated execution scanned 4,708 targets and deployed 32 web shells in 61 seconds.

Who is CyberXero and what is their goal?

CyberXero is assessed as a financially motivated initial access broker. The group breaks into networks to sell access to other criminals rather than using it directly. While the global pipeline targets WordPress and e-commerce sites for payment data, the curated campaign focuses on deep reconnaissance of Ukrainian utilities, including a heating provider in Kharkiv, a city on an active war front.

What is the current status of the threat?

SOCRadar stated that the campaign remains ongoing. The actor used 51 AI agents to support the attacks, combining commodity offensive tooling with an AI orchestration layer. The exposed logs revealed every planned target, refused task, and workaround attempt, effectively acting as a keylogger for the attacker’s own operations. No fix for the underlying infrastructure misconfiguration has been confirmed yet.

Background: AI agents

AI agents are software systems that perceive their environment, make decisions, and execute actions to achieve specific goals. They move beyond generating text to performing tasks like coding, scheduling, or data analysis. You must treat them as active participants in your network, not passive tools.

Read the full guide: AI Agents Examples: How Autonomous Software Changes Security

What to do and how to stay safe: CyberXero

• Audit web servers for open directories that allow directory listing or unauthorized access to sensitive files.
• Rotate any API tokens, credentials, or keys that may have been stored in plaintext in exposed directories.
• Monitor WordPress sites for signs of unauthorized admin accounts or web shells, especially if using the Support Board plugin.
• Review access logs for unusual scanning activity or rapid automated requests that could indicate exploitation attempts.

Step-by-step guide: AI Usage Policies at Work: How to Prevent Data Leakage

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

Frequently asked questions

What CVEs are associated with this campaign?

The campaign exploits CVE-2026-63030, CVE-2026-60137, and CVE-2026-4815, an SQL injection vulnerability in the Support Board WordPress plugin.

How many records were stolen from Ukrainian organizations?

More than 628,000 records tied to Ukrainian individuals were confirmed to be in the actor’s possession, according to SOCRadar.

What AI tool did CyberXero use?

The threat actor used Anthropic's Claude Code as part of an AI orchestration layer to support the cyberattacks.