Underground Forum Advertises 'Uncensored' AI Tool Built for Cybercrime

Researchers from Sophos's Counter Threat Unit have identified a forum user advertising a subscription AI service called Luciferus on the underground marketplace Exploit. The seller claims the tool answers requests without moral or ethical limits and is based on a custom model with 120 billion parameters, though these claims have not been independently verified.

While the exact technology behind Luciferus is unconfirmed, researchers believe with low confidence it may be built on Qwen, an open-source AI model family developed by Alibaba. This follows a pattern seen previously where criminals sold 'jailbroken' versions of mainstream tools like ChatGPT and Claude with safety restrictions stripped out. Tools like Luciferus go further by being built or configured from the start without safeguards, giving criminal users more control and the ability to tailor the AI for malicious purposes such as generating phishing content, malware code, or scam scripts.

Such 'proprietary model' claims should be treated with scepticism, as many of these services are likely repackaged open-source models with custom prompts rather than genuinely new AI systems built from scratch. Regardless of the underlying technology, the emergence of purpose-built uncensored AI tools signals that cybercriminals are increasingly using AI to scale up attacks like phishing and social engineering.