Researchers Identified Uncensored AI Service Luciferus
The subscription-based model reportedly generates malicious code without standard safety filters.
Updated on Sept. 26, 2026 in Artificial Intelligence
Live Poll
Is the growth of unregulated AI services making it too easy for individuals to conduct cybercrime?
Sophos researchers reported that an AI service named Luciferus appeared on the Exploit underground forum on August 24, 2026. The service is marketed as a proprietary tool with 120 billion parameters capable of generating restricted content.
Why it matters
The development represents a trend of threat actors packaging large language models for cybercrime, offering a stable alternative to mainstream jailbroken tools. It highlights how criminal ecosystems are now adopting standard commercial subscription models to distribute malicious capabilities.
The Luciferus AI service is advertised as a proprietary model featuring 120 billion parameters. Sophos researchers observed the service's 'Luciferus Junior' variant generating code for a remote access trojan, though the underlying architecture remains unverified.
The players
Sophos X-Ops
A cybersecurity research unit that tracks threat actor activities, malware distribution, and emerging defensive strategies.
Optimus_Prime
An alias for an Exploit forum user who advertised the availability of the Luciferus AI subscription service.
The details
Luciferus operates using a tier-based subscription structure similar to established phishing and malware kit markets. The service specifically advertises the ability to bypass moral and ethical restrictions common in mainstream models, allowing users to generate restricted content. During observations, the model successfully drafted a Python-based remote access trojan—a type of malware that provides unauthorized control over a remote computer system.
Timeline
August 24, 2026: Luciferus was officially advertised on the Exploit forum.
The Tech Race
This service follows the documented shift of threat actors commoditizing AI-based malware tools for the cybercrime underground. It marks a transition from bespoke jailbreaking to the sale of dedicated, uncensored large language models as a service.
This research primarily affects security analysts and organizations monitoring for new vectors of automated malware generation. There is currently no public access to the service for legitimate users, as it is being marketed exclusively within underground forums.
The takeaway
The rise of specialized, subscription-based AI models for criminal use suggests that automated malware generation is becoming a scalable industry service. Security teams should monitor threat intelligence feeds for signatures linked to Luciferus-generated code as indicators of novel attack vectors.
Further reading
For broader trends in secure AI development, see Artificial Intelligence.
Source note: This article includes information reported by Upgrade Magazine.
Live Poll
Is the growth of unregulated AI services making it too easy for individuals to conduct cybercrime?
