Skip to content
Today
Sections

Security Independently confirmed2 publishers 3 min read Published Updated

Luciferus sells uncensored malware generation on the Exploit forum starting at $35 a month

Sophos's Counter Threat Unit found the advertisement on August 24. The cheapest tier answered a direct request for a Python remote access trojan with source code, and the service runs on a local model no provider can patch.

The Watch · Security desk

Send a correction

What happened

  • Sophos's Counter Threat Unit saw a persona called Optimus_Prime advertise Luciferus, an uncensored AI subscription service, on the Exploit underground forum on August 24, 2026.
  • The advertisement says the service answers requests without moral or ethical restrictions and claims it is based on a proprietary model with 120 billion parameters.
  • The forum ad prices three monthly tiers, Inquisitor at $35, Archdeviel at $55 and Prince of Darkness at $75, plus an Individual Embodiment VIP level quoted on request.
  • The Luciferus website carries different tier names and lower prices, Junior at $22, Middle at $34.75 and Pro at $47.14, and omits the VIP option entirely.

Why it matters

  • What it limits constraint Provider-side patching has no effect on this buyer: nothing OpenAI or Anthropic ships degrades a locally hosted uncensored model, so removal depends on the seller's infrastructure.
  • What it costs cost At $264 a year for the cheapest tier, malware code on demand sits inside the budget of an actor with no development skill, and the seller collects recurring revenue whether or not the output ever works.
  • What's now possible capability The Individual Embodiment tier offers dedicated compute and models trained on the buyer's own data. A crew could tune output to its own targets on hardware shared with no other customer.
  • What doesn't add up contradiction Two price lists for one product, flagged by Sophos without explanation, leave buyers and trackers unable to say which storefront is the operating one.

Divide each website price by its forum counterpart and the same figure comes back three times: $22 against $35 is 0.63, $34.75 against $55 is 0.632, $47.14 against $75 is 0.629 [21]. The website list looks like the forum list run through one multiplier. Sophos flagged the mismatch in names and prices and did not explain it [9].

For a defender the difference that counts sits upstream of the price. According to Help Net Security's account of the Sophos research, jailbroken builds of ChatGPT or Claude lose their restrictions the moment a vendor patches them, while tools built without safeguards from the start give sellers a longer shelf life [15]. Sophos describes services of this kind as designed or configured without safeguards from the outset, which the researchers tie to greater control, persistence, and the ability to tailor capabilities to specific user communities [14]. "Luciferus appears to be a more stable option that relies on an uncensored local LLM instead of jailbreaking a mainstream LLM provider," CTU researchers wrote [13]. A patch shipped by a mainstream provider costs this seller nothing.

Everything the advertisement says about the model itself is unverified. CTU did not check the architecture, the parameter count, the performance or the privacy claims, and it puts only low confidence on its own assessment that Alibaba's Qwen family is underneath [4][5]. The researchers were blunt about the proprietary-model claim: "Proprietary model claims can be misleading, as many of the 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 LLM requires significant expertise, data, and computing resources," they wrote [16].

The test result is narrow. The Junior tier answered a prompt for a simple Python RAT with a Russian-language description of its networking and command execution functions, then source code, which Sophos says supports the ad's claim that the service will respond to overt malware requests [10]. CTU stopped short of running the code or checking whether it was complete [11]. Sophos did not report how many subscribers the service has.

Optimus_Prime has a thin record on the forum: 21 posts in the 139 days between joining on April 18 and September 4, about one a week [2][22]. The category around the persona is busier. WormGPT and FraudGPT were sold the same way, as unrestricted ChatGPT alternatives for phishing emails, BEC lures, malicious scripts and malware code [17], and CTU also tracks brokered access to legitimate AI platforms, shared API keys, prompt-engineering services and multi-model packages [18]. Posts advertising AI-enabled services are increasing, alongside dedicated AI discussion channels and recruitment aimed at AI specialists [19].

At $22 a month the cheapest tier runs $264 a year; the top forum tier at $75 runs $900 [23]. "These developments suggest that AI is 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 wrote [20].

What to watch

  • Whether anyone obtains a sample and tests the 120-billion-parameter claim or the low-confidence Qwen attribution.
  • Whether the website storefront and the forum tier list converge, or one of them goes dark. Either outcome would show which channel is live.
  • Whether other Exploit sellers copy the Individual Embodiment offer of dedicated compute and training on buyer-supplied data.

Clarity's read

What the record supports and how the coverage leans. The claims behind it follow.

Reality

Evidence55
Adoption
Not scored
Hype gap+30
Incentives55
Confidence60
Why these scores

Claim ledger

Ranked by verification strength, evidence, and original report placement.

  1. [1]

    On August 24, 2026, Sophos Counter Threat Unit researchers observed an Exploit underground forum persona named "Optimus_Prime" advertising an uncensored AI subscription service named Luciferus.

    Reported SupportedView cited source
  2. [2]

    The Optimus_Prime persona joined Exploit on April 18, displays a "coding / coder" activity label, and as of September 4 had published 21 posts on the forum.

    Reported SupportedView cited source
  3. [3]

    The advertisement describes Luciferus as an AI system that answers requests without moral or ethical restrictions and claims it is based on a proprietary model with "120 billion parameters".

    Reported SupportedView cited source
  4. [4]

    CTU researchers did not independently verify the model architecture, parameter count, performance, privacy claims, or advertised capabilities.

    Reported SupportedView cited source
  5. [5]

    CTU researchers assess with low confidence that Luciferus is built on Qwen, a family of large language models developed by Alibaba.

    Reported SupportedView cited source
  6. [6]

    The forum advertisement lists three monthly subscription tiers: Inquisitor ($35), Archdeviel [sic] ($55), and Prince of Darkness ($75).

    Reported SupportedView cited source
  7. [7]

    The ad also refers to an "Individual Embodiment" VIP level providing a personal AI model deployed separately for the buyer's project, custom models trained on the buyer's own data and specific tasks, dedicated computing power not shared with other users, and full control over the context window and response temperature, priced according to user requirements.

    Reported SupportedView cited source
  8. [8]

    The Luciferus website does not mention the VIP option and lists different tier names and prices: Junior ($22), Middle ($34.75), and Pro ($47.14).

    Reported SupportedView cited source
  9. [9]

    Sophos flagged the discrepancy between the forum and website tier names and prices without explanation.

    Reported SupportedView cited source
  10. [10]

    The Luciferus Junior tier's response to a "simple RAT in python" prompt gave a Russian-language introduction describing a simple Python remote access trojan and its networking and command execution functionality before listing source code, results that CTU said support the advertisement's claim that the service will respond to overt requests for malware development.

    Reported SupportedView cited source
  11. [11]

    CTU researchers did not execute, test, or assess the completeness of the generated code.

    Reported SupportedView cited source
  12. [12]

    CTU researchers previously observed threat actors advertising "jailbroken" versions of ChatGPT and Claude in which ethical constraints were removed so they could be used for malicious purposes.

    Reported SupportedView cited source
  13. [13]

    "Luciferus appears to be a more stable option that relies on an uncensored local LLM instead of jailbreaking a mainstream LLM provider," CTU researchers wrote.

    Reported SupportedView cited source
  14. [14]

    Sophos says "proprietary models" like Luciferus are designed or configured without safeguards from the outset to allow greater control, persistence, and the ability to tailor capabilities to specific user communities, including underground forums.

    Reported SupportedView cited source
  15. [15]

    Help Net Security's account of the Sophos research states that unlike jailbroken versions of ChatGPT or Claude, which lose their restrictions once a vendor patches them, tools like Luciferus are built without safeguards from the start, giving sellers more control and a longer shelf life.

    Reported SupportedView cited source
  16. [16]

    "Proprietary model claims can be misleading, as many of the 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 LLM requires significant expertise, data, and computing resources," researchers wrote.

    Reported SupportedView cited source
  17. [17]

    WormGPT and FraudGPT were marketed on underground forums as unrestricted alternatives to ChatGPT capable of generating phishing emails, business email compromise lures, malicious scripts, and malware code.

    Reported SupportedView cited source
  18. [18]

    CTU researchers have observed threat actors selling brokered access to legitimate AI platforms, sharing API keys, offering AI prompt-engineering services, and advertising access to multiple AI models as a packaged service.

    Reported SupportedView cited source
  19. [19]

    CTU researchers have identified a steady increase in posts advertising AI-enabled services, dedicated AI discussion channels, and recruitment efforts seeking AI specialists to support cybercriminal operations.

    Reported SupportedView cited source
  20. [20]

    "These developments suggest that AI is 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 concluded.

    Reported SupportedView cited source
  21. [21]

    Each website tier price is about 63% of the corresponding forum tier price: 22/35 = 0.629, 34.75/55 = 0.632, 47.14/75 = 0.629.

    Derived Supported
  22. [22]

    April 18 to September 4 is 139 days, so 21 posts works out to about one post a week.

    Derived Supported
  23. [23]

    The $22 website tier costs $264 a year and the $75 forum tier costs $900 a year.

    Derived Supported

Sources

2 independent publishers whose own reporting we read for this story.

  1. helpnetsecurity.com

    1 article · September 15, 2026

    Uncensored AI sold on hacking forum as alternative to ChatGPT and Claude jailbreaks
  2. nakedsecurity.sophos.com

    1 article · September 13, 2026

    Devil’s advocate? Uncensored Luciferus AI service advertised underground

Topics and entities

Topics

  • LLM jailbreaking
  • Underground forums
  • AI-enabled cybercrime

Entities