Threat Actors attempt to use Xanthorox AI Tool to Generate Different Malicious Code

Du'aine Davis

For awareness

• On 11 November, Trend Micro tested Xanthorox, a criminal-focused LLM advertised since early 2025, for its ability to generate malicious code, assist with OSINT, and enable phishing. Despite its hype, the tool fell short: it cannot browse the internet, access darknet forums, or retrieve real-time vulnerability data. Its fine-tuning prioritises removing guardrails and adding a persona rather than improving technical capability.
• Xanthorox operates as a jailbreak-as-a-service model, likely leveraging Google’s Gemini Pro infrastructure despite claims of self-hosting. It offers features like Agentex for compiling executable malware, subscription tiers payable in cryptocurrency, and detailed coding assistance for shellcode and obfuscation. While its lack of RAG and browsing limits reconnaissance, its ability to produce functional malicious code reflects a growing trend of threat actors adopting LLMs for cybercrime.
• Organisations should monitor for behaviours linked to AI-assisted attacks, such as rapid generation of obfuscated scripts or custom malware. Microsoft Defender users should enable advanced threat protection, apply strict egress filtering, and watch for anomalous code execution patterns. Ingress/egress filtering to detect malicious prompts and outputs from AI tools is also recommended.

The Devil Reviews Xanthorox: A Criminal-Focused Analysis of the Latest Malicious LLM Offering | Tre…