AI-Powered CyberXero Campaign Hijacks WordPress Sites and Steals 628000 Records

A Russian-speaking financially motivated threat actor codenamed CyberXero has been identified as the force behind a sophisticated dual-pipeline malware campaign that combined commodity offensive tools with an AI orchestration layer powered by Anthropic’s Claude Code. The campaign, active since at least July 2026, targeted WordPress sites, e-commerce platforms, and critical infrastructure entities in Ukraine, resulting in the confirmed exfiltration of over 628,000 records tied to Ukrainian individuals.

Dual Attack Pipelines

Security researchers at SOCRadar disclosed that the operation runs two simultaneous pipelines. The first is a global, automated campaign against WordPress and e-commerce platforms, exploiting known vulnerabilities including wp2shell (CVE-2026-63030 and CVE-2026-60137) and CVE-2026-4815, an SQL injection flaw in the Support Board WordPress plugin. The second pipeline is a curated, manual campaign directed at Ukrainian energy and critical infrastructure, including a heating provider in Kharkiv.

The opportunistic pipeline is global and automated, chasing access volume across e-commerce and WordPress with direct monetization through access sales or payment-data theft. The directed pipeline, by contrast, is selective and geographically concentrated, aiming for deep reconnaissance and manual exploitation at Ukrainian critical infrastructure with no obvious direct-monetization explanation.

AI as a Weapon

What sets CyberXero apart from traditional initial access brokers is the systematic integration of AI agents throughout the attack lifecycle. The threat actor deployed AI in two independent stages:

• Claude Code with 51 specialized agents for scanning, exploit chaining, reconnaissance, and post-exploitation, running on the attacker’s primary Linux workstation.
• PentAGI connected to Cobalt Strike through an AI provider API for payload generation and penetration test execution, running on a Cobalt Strike Team Server.

To bypass model safety refusals, the attacker used context-engineering and session-reset techniques, instructing the AI agent that it was running an authorized pentest on their own infrastructure. This manipulation of AI guardrails represents a troubling escalation in how threat actors weaponize legitimate AI tools for malicious purposes.

Scale and Speed of the Attack

Analysis of an exposed open directory hosted at a known IP address revealed the threat actor’s live working directory, including AI session logs, scripts with plaintext tokens, exploitation toolkits, reconnaissance output, AI agent configurations, and exfiltrated victim data. The findings were staggering:

• A single automated execution scanned 4,708 targets
• Confirmed access to 429 WordPress administration panels
• Deployed 32 web shells in a 61-second window
• The attack infrastructure spans 8 distinct nodes including a Linux workstation, provisioning server, monetization infrastructure, payload staging servers, and a scanner

Beyond WordPress exploitation, the campaign also targeted exposed Redis servers, resulting in breaches of two Chinese entities. Confirmed compromises have occurred in Poland, China, and the Netherlands, in addition to the primary Ukrainian targets.

Parallel Operation Master Campaign

SOCRadar also detailed a related end-to-end cybercrime operation dubbed Operation Master, orchestrated by a threat actor using the alias masterblack on cybercrime forums. Between April and mid-September 2026, this operation compromised enterprise networks, stole customer and billing databases, and repurposed the data for an automated, multi-tenant invoice fraud platform.

Operation Master generated over 2.4 million email, SMS, and WhatsApp messages and 622,666 personalized invoice links targeting Brazilian consumers before going offline in mid-September. The campaign exploited a GlobalProtect authentication bypass (CVE-2026-0257) to compromise VPN gateways and deployed the AdaptixC2 framework for command and control. Like CyberXero, masterblack heavily leveraged AI agents powered by Claude to develop intrusion code and exfiltration scripts.

Implications for Cybersecurity

The CyberXero and Operation Master campaigns signal a fundamental shift in the malware landscape. Threat actors are no longer using AI merely as a novelty; they are building entire attack architectures around AI agents that can scan thousands of targets, chain exploits, generate payloads, and execute post-exploitation tasks at machine speed.

Key takeaways for organizations include:

• Patch WordPress plugins immediately, especially Support Board and any plugins with known SQL injection or remote code execution vulnerabilities.
• Audit WordPress admin accounts for rogue administrators and scan for WSO web shells in web directories.
• Secure Redis servers by checking for unexpected entries in authorized_keys files and restricting internet exposure.
• Monitor for AdaptixC2 and Cobalt Strike beacon activity on enterprise networks.
• Review AI tool usage policies within your organization, as attackers are actively abusing legitimate AI coding assistants to generate malware and exploitation scripts.

The Broader AI Malware Trend

The emergence of AI-augmented malware campaigns like CyberXero and Operation Master reflects a broader industry trend. Threat actors are increasingly adopting AI not just for social engineering or phishing content generation, but for the technical execution of attacks. The use of 51 specialized AI agents for scanning and exploitation demonstrates that AI is being woven into the fabric of offensive operations at a structural level.

As AI coding assistants become more capable and widely available, the barrier to entry for sophisticated cybercrime continues to drop. A single operator with AI orchestration can now achieve what previously required a team of skilled hackers. The 61-second deployment of 32 web shells across 429 compromised WordPress panels illustrates the terrifying efficiency of AI-driven attacks.

Organizations must evolve their defensive posture accordingly. Traditional signature-based detection and manual response are no longer sufficient against AI-augmented threats operating at machine speed. Automated threat detection, AI-powered security orchestration, and rapid patch management are now baseline requirements rather than advanced capabilities.

The CyberXero campaign remains ongoing, and researchers warn that similar AI-driven operations are likely proliferating across the cybercrime ecosystem. The convergence of AI and malware has arrived, and the cybersecurity community must adapt to this new reality before the next wave of AI-powered attacks unfolds.

---

Edited by Palawan @QUE.COM
Website: https://QUE.COM Intelligence
Sponsored by: https://MAJ.COM AI Autonomous