Cybercriminals are increasingly using ChatGPT and Claude for malicious purposes, although they still have to ‘trick’ the well-known OpenAI and Anthropic assistants into carrying out questionable actions.
But what if a similar tool gave threat actors the full power of generative AI for malicious activities, without any ethical constraints?
Something similar already exists. It is called Luciferus and is beginning to appear on forums aimed at black-hat hackers.
Cybersecurity company Sophos, through its Counter Threat Unit (CTU), has found an advertisement on one such platform for the demonically named AI tool, which is designed to do harm.
According to the post, the tool runs on a proprietary model with 120 billion parameters.
The author describes himself as a programmer and coder and uses the alias Optimus_Prime, after the leader of the Transformers. His forum account was registered on 18 April, and he has since posted around 20 messages.
Modified or ‘jailbroken’ versions of ChatGPT and Claude had already been discovered, allowing the well-known LLMs to be used maliciously without restrictions. However, this new ‘weapon’ has been built from the ground up without safeguards. This could give users greater control and allow them to keep using it for much longer.
Perhaps not so ‘proprietary’
Sophos researchers question whether the tool was developed as its creator claims.
“Claims of proprietary models can be misleading, as many services are likely based on fine-tuned open-source models, custom system prompts or orchestration layers rather than entirely new foundation models. Training a genuinely novel machine-learning model requires extensive expertise, data and computing resources”, they say.
The forum advertisement says that Luciferus offers three subscription tiers: Inquisitor costs $35 a month, Archdevil $55 and Prince of Darkness $75.
There is also a separate tier called ‘Individual Incarnation’. This consists of a customised private model trained on the customer’s own data, with dedicated computing power and control over the context-window size and response temperature. Pricing depends on the buyer’s requirements.
“The emergence of Luciferus aligns with a broader trend in which cybercriminals are increasingly commercialising AI through underground forums, Telegram channels and cybercrime marketplaces”, the researchers say.
“These developments suggest that AI is becoming an increasingly accessible component of the cybercrime ecosystem, lowering barriers to entry and enabling less technically skilled actors to access capabilities that were previously limited to experienced developers”, Sophos concludes.
Click Here For The Original Source.
