Cybersecurity researchers have uncovered an AI-assisted cybercrime campaign linked to a threat actor known as CyberXero, which allegedly stole more than 628,000 records connected to Ukrainian individuals. The operation combined AI coding agents, conventional hacking tools and automated exploitation techniques to target websites, e-commerce platforms and critical infrastructure.
Reported by The Hacker News on October 12, 2026, the campaign was documented by cybersecurity firm SOCRadar. Researchers attributed the activity to a Russian-speaking, financially motivated initial access broker believed to have been active since at least July 2026. An initial access broker compromises systems and may sell access to other cybercriminals. <Cite refs={["turn324665news0","turn324665news2"]}/>
How the Attack Used AI Agents
Investigators found that CyberXero used Anthropic's Claude Code alongside other offensive security tools to automate different stages of its operations.
According to SOCRadar, the attacker configured a cluster of 51 specialised AI agents to assist with scanning, reconnaissance, vulnerability exploitation and post-exploitation activities. The operation also connected PentAGI, an open-source AI penetration-testing framework, with Cobalt Strike infrastructure.
One automated execution reportedly scanned 4,708 targets, gained access to 429 WordPress administration panels and deployed 32 web shells within 61 seconds. Web shells can provide attackers with persistent remote access to compromised servers. <Cite refs={["turn324665news0","turn324665news2"]}/>
More Than 628,000 Records Stolen
Researchers confirmed data exfiltration from four Ukrainian organisations, with more than 628,000 records linked to Ukrainian individuals found in the attacker's possession.
The campaign involved two parallel operations:
Global website attacks: Automated targeting of WordPress and e-commerce platforms to obtain system access and potentially steal payment-related information.
Targeted infrastructure reconnaissance: More selective investigations of Ukrainian energy and utility organisations, including systems associated with critical infrastructure.
Researchers also identified compromises involving organisations in Poland, China and the Netherlands. The investigation remains ongoing, and the full extent of the damage has not been established. <Cite refs={["turn324665news0","turn324665news2"]}/>
AI Safety Controls Faced Attempts to Be Circumvented
The exposed attacker logs revealed attempts to work around AI safety restrictions. When the system refused certain requests, the operator reportedly restarted sessions and framed the activities as authorised security testing.
However, the logs also documented refusals that remained in place. The AI system declined several requests involving backdoors, firewall disabling, lateral movement and web-shell deployment.
These findings illustrate a key cybersecurity challenge: AI tools designed to improve legitimate software development and security testing can also be misused when controlled by malicious operators. The incident does not mean that Claude Code itself was breached; rather, researchers reported that the attacker used the tool as part of a broader offensive setup. <Cite refs={["turn324665news0","turn324665news2"]}/>
What Businesses Should Do to Protect Their Systems
The incident highlights the importance of securing website infrastructure and AI-enabled development environments. Organisations should prioritise the following measures:
Patch vulnerable software: Update WordPress, e-commerce platforms and plugins, and review vendor security advisories.
Audit administrator accounts: Investigate unfamiliar users, unexpected privileges and suspicious login activity.
Inspect for malicious files: Check for unauthorised web shells and unexpected changes to website files.
Secure exposed services: Restrict public access to databases, Redis instances and other sensitive infrastructure.
Protect credentials: Rotate exposed tokens and review SSH keys, API keys and other authentication mechanisms.
Monitor AI tools: Apply access controls, maintain audit logs and investigate unusual automated activity.
These measures can help reduce the risk of unauthorised access and improve incident detection.
What This Means for the Cybersecurity Industry
The CyberXero campaign highlights how AI agents can accelerate cyber operations when combined with existing exploitation tools. Tasks that previously required substantial manual effort can potentially be coordinated across multiple automated agents.
For businesses, this means security strategies must account not only for traditional malware and vulnerabilities but also for AI-enabled automation, exposed credentials and misuse of developer tools.
Key Takeaway:
CyberXero allegedly used Claude Code and other offensive tools in a campaign that compromised hundreds of WordPress administration panels and resulted in the theft of more than 628,000 Ukrainian records. The findings reinforce the need for timely software updates, stronger access controls and continuous monitoring of both traditional infrastructure and AI-powered tools.