Underground Forum Advert Pushes 'Luciferus', an AI Service Sold Without Safety Limits

Sophos Counter Threat Unit (CTU) researchers reported that on 24 August 2026 a persona called "Optimus_Prime" advertised an uncensored AI subscription service named Luciferus on the Exploit underground forum. The persona joined the forum on 18 April and had published 21 posts as of 4 September. The advert says the service answers requests without moral or ethical restrictions and claims it runs on a proprietary model with "120 billion parameters".

The researchers did not independently verify the model design, size, performance, privacy claims or advertised capabilities. They assess with low confidence that Luciferus is built on Qwen, a family of large language models developed by Alibaba. They also note that "proprietary model" claims can be misleading. Many such services are likely based on fine-tuned open-source models, custom system prompts or orchestration layers, because training a genuinely new model takes significant expertise, data and computing resources.

CTU has previously seen threat actors advertise "jailbroken" versions of ChatGPT and Claude with their ethical limits removed. Services like Luciferus differ because they are designed or configured without safeguards from the start. According to the researchers, this gives sellers more control and persistence, and lets them tailor features to specific user communities such as underground forums. This summary covers only the opening of the Sophos report, not the full analysis.