Unrestricted AI Available on Dark Web as Alternative to ChatGPT and Claude Jailbreaks

Post Views: 1

Uncensored AI subscription service Luciferus promoted on dark web forum as alternative to ChatGPT and Claude jailbreaks.

Discovery of Luciferus on the Dark Web Forum

According to findings by cybersecurity researchers, a new AI subscription service named Luciferus is being promoted on a dark web forum as a substitute for circumventing restrictions in ChatGPT or Claude. The Counter Threat Unit (CTU) detected the advertisement on August 24 on the Exploit forum, posted under the username Optimus_Prime. The account was created on April 18 and has generated 21 messages, with a profile description indicating a focus on coding.

Forum Post Details

The forum post characterizes Luciferus as a system designed to respond to queries without ethical or moral constraints. It claims the platform operates on a proprietary model with 120 billion parameters, though Sophos could not validate this assertion or other performance and privacy claims outlined in the advertisement.

Model Claims and Validation

Researchers assessed with low confidence that Luciferus may be based on Qwen, an open-source large language model family developed by Alibaba. They noted that such proprietary model claims often lack transparency. Many similar services likely rely on fine-tuned open-source models, custom system prompts, or orchestration layers rather than entirely new foundation models.

Technical Challenges of Developing New Models

Developing a novel large language model requires substantial expertise, data, and computational resources, as highlighted by the researchers.

Pricing Structure and Subscription Tiers

Subscription pricing varies between the forum and the service’s official website. The forum ad outlines three tiers: Inquisitor at $35 monthly, Archdevil at $55, and Prince of Darkness at $75. A separate Individual Embodiment tier offers a private model tailored for specific projects, including custom data training, dedicated computing infrastructure, and adjustable parameters like context window size and response temperature, with costs determined by individual requirements.

Website Pricing Discrepancies

The Luciferus website lists different pricing structures, including Junior at $22, Middle at $34.75, and Pro at $47.14, but omits the VIP option mentioned in the forum post, a discrepancy noted by Sophos without further clarification.

Testing and Capabilities

Testing conducted by Sophos involved requesting a basic remote access trojan written in Python. The Junior tier provided a Russian-language explanation of the malware’s networking and command execution functions, followed by source code. While the researchers did not execute or verify the code, the response aligns with the advertisement’s claim that the service fulfills direct malware-related requests.

Evolving Cybercrime Landscape

The emergence of Luciferus reflects a growing trend of threat actors commercializing AI tools through underground forums, Telegram channels, and cybercriminal marketplaces, as observed by Sophos. The service is positioned as a more stable alternative to jailbroken mainstream models, which lose their unrestricted capabilities after vendor patches.

Implications for Cybersecurity

By operating without built-in safeguards, Luciferus offers sellers greater control and longevity. These developments underscore the increasing integration of AI into the cybercrime ecosystem, lowering technical barriers for less skilled actors to access advanced capabilities previously limited to experienced developers.

The proliferation of such services highlights the evolving landscape of AI-driven threats and the need for enhanced monitoring of underground platforms.