Sophos Uncovers Luciferus, an Underground Uncensored AI Service Sold to Bypass Every Content Restriction

Sophos Uncovers Luciferus, an Underground Uncensored AI Service Sold to Bypass Every Content Restriction

An uncensored AI subscription service called Luciferus is being marketed on underground forums as a way to bypass the guardrails of mainstream models — and it has quietly been signing up users since late August. First spotted by Sophos's Counter Threat Unit on August 24, 2026, the service is being sold on the Exploit forum by a persona calling themselves "Optimus_Prime" and promises to answer requests without moral or ethical restrictions.

While cybersecurity coverage has focused on Luciferus's ability to generate malware code — researchers successfully prompted it to produce a Python remote-access trojan without safeguards — the service has a broader implication for the NSFW AI space. Any model sold as fully uncensored inevitably sidelines adult-content filters, giving users a turnkey alternative to jailbreaking ChatGPT, Claude, or Gemini for sexual roleplay or explicit content generation.

What Luciferus Claims to Be

The seller claims Luciferus runs on a proprietary 120-billion-parameter model. Sophos analysts assessed this claim with low confidence and noted the model may instead be built on Alibaba's open-weight Qwen family, which has previously been used to create uncensored fine-tunes. The service is offered on a subscription basis with tiers named Inquisitor ($35/month), Archdeviel ($55/month), and Prince of Darkness ($75/month) — though a publicly accessible landing page later surfaced showing different tier names and lower prices, suggesting the offering is still being tuned.

As of September 4, the same forum persona had published 21 posts advertising and supporting the service. Security Point Break found the public landing page on September 15 offering a $35/month plan with four free prompts requiring no credit card, indicating the operator is actively expanding distribution and lowering the barrier to entry.

Why This Matters for the NSFW AI Space

Luciferus is the latest example of a growing underground market for models with zero content moderation. The service sits alongside other uncensored offerings such as Abliteration.ai — a startup covered by TechCrunch in early September that stripped refusal mechanisms from open-weight models and offered them via API. Unlike mainstream platforms that are tightening NSFW policies — OpenAI, Google, Anthropic, and Meta all maintain strict blocks on sexual content in their consumer products — Luciferus represents the opposite trend: a supply chain for unrestricted output that operates entirely outside corporate policy.

For companion and NSFW AI users, this means the "uncensored AI" ecosystem is fragmenting. Legitimate NSFW platforms like Candy AI, SpicyChat, and CrushOn operate within legal boundaries with gated adult content and age verification. Underground services like Luciferus operate with no guardrails at all — a distinction that matters for safety, legality, and staying power. The latter can disappear overnight if law enforcement moves on the operator, as seen when the Manhattan DA recently seized domain names for 12 AI deepfake porn sites.

The Enforcement Backdrop

Luciferus is surfacing at a time when both law enforcement and regulators are paying closer attention to unmoderated AI. The same week Sophos publicized its findings, New York's Manhattan District Attorney seized domain names for a dozen AI deepfae prn sites, as reported by The Record. In the UK, the Advertising Standards Authority banned five AI companion and nudify app ads on Meta platforms. California's SB 243 companion chatbot law is now enforceable, and half of US states require age verification for adult content.