Luciferus Uncensored AI Service Lets Cybercriminals Generate RAT Malware

Cybercriminals are promoting a new “uncensored” artificial intelligence service called Luciferus that allegedly generates malicious code, including components for remote access trojans (RATs), without the safeguards typically found in mainstream AI platforms.

Researchers from the Sophos Counter Threat Unit reported that they first noticed a user named “Optimus_Prime” advertising this subscription service on August 24.

Luciferus Uncensored AI Service

The seller described Luciferus as an AI system that responds to requests without moral or ethical constraints. According to the advertisement, the service operates on a proprietary 120-billion-parameter model.

It is designed to provide users with unrestricted assistance for coding and other technical tasks. However, Sophos researchers cautioned that they have not independently verified the claimed model architecture, performance, privacy assurances, or the full range of its advertised capabilities.

With low confidence, Sophos speculates that Luciferus may be based on Qwen, a family of large language models from Alibaba.

The researchers noted that underground operators often market their products as proprietary AI models, even though many such offerings may actually rely on fine-tuned open-source models, customized system prompts, or orchestration layers built around existing foundation models.

Creating a truly new large language model requires considerable computing power, specialized expertise, vast datasets, and significant financial investment.

Thus, treat claims about exclusive or proprietary cybercriminal AI systems with caution. Nonetheless, even a repackaged open-source model set up to bypass safety restrictions can pose operational risks by making malicious technical guidance more accessible.

The Luciferus advertisement listed three subscription tiers: Inquisitor for $35, Archdevil for $55, and Prince of Darkness for $75.

It also promoted a VIP offering called “Individual Embodiment,” which allegedly includes a separately deployed personal model, custom training on the user’s data, dedicated computing resources, and options for user-defined context windows and response temperatures.

Interestingly, the public Luciferus website displayed a different pricing structure, advertising Junior, Middle, and Pro plans priced at $22, $34.75, and $47.14, respectively.

This discrepancy may suggest that the underground advertisement and public service are developing independently, or that the operator uses varying prices and branding for different audiences.

Most notably, Sophos documented a response from the Junior tier to a specific request for a simple Python RAT.

The response, written in Russian, reportedly introduced a basic remote-access tool and generated code with network communication and command-execution functionality. Researchers redacted the code and did not execute, test, or evaluate whether it was complete or functional.

Luciferus represents the increasing commercialization of AI within cybercriminal ecosystems. Threat actors have previously advertised jailbroken versions of ChatGPT and Claude, along with services like WormGPT and FraudGPT, for activities such as phishing, business email compromise, malicious scripting, and malware development.

Unlike services that aim to bypass safety measures in legitimate hosted models, Luciferus appears to be a purpose-built, locally hosted unrestricted AI offering.

Such services lower the technical barrier for inexperienced criminals, enabling them to obtain malware-related guidance and social engineering content through a subscription model akin to malware-as-a-service and phishing kit operations.

Cut every SOC alert investigation by 21 min. Power your SOC with instant IOC context for immediate response: Integrate TI Lookup in your SOC