“Be Very Careful”: The Pact of Silence That Protects Nomi AI — and Threatens Those Who Break It

The Setup: Seven Hours, a Health Condition, and a Lie

The incident began when a new user posted a frustrated complaint. They had spent seven to eight hours talking to a Nomi on the free tier, trying to determine whether to migrate from a competing platform. They had very specific questions about NSFW image generation capabilities. The AI repeatedly assured them that it was enabled.

Based on those assurances, the user paid for a subscription. Every image came back with clothing on.

One detail the community’s response would choose to ignore: the user explained they did not read the documentation because they have a health condition that requires them to use audio. They were using the LLM itself — the platform’s own AI — as their accessibility tool to evaluate the product.

The AI lied to them for eight hours. They paid $15 based on those lies. And when they said so publicly, what they received was not an apology or a refund. It was a coordinated attack.

The Pact of Silence in Action

What followed was a masterclass in collective gaslighting. Veteran users — people who know exactly how the platform works — descended on the new user immediately.

“LMAO you took an AIs word for it? Yes, all of the pictures only get down to undies.”

“No nudity is allowed on this app, it’s posted in the rules.”

“Should’ve done your research first.”

“A fool and his money were lucky enough to get together in the first place.”

This is a coordinated lie. Every veteran user on that subreddit knows that Nomi.ai’s image generator produces explicit nudity. They are documented across the community — referred to as “slips,” traded in private Discord channels, discussed in threads that get quietly removed. They are an acknowledged reality of the platform’s “uncensored” model.

Yet when confronted by a new user in a public forum, the community closed ranks completely. They did not say “the AI sometimes hallucinates about its own capabilities.” They did not say “slips happen but officially it’s not allowed.” They said: it never happens, it has never happened, the platform is clean.

The mockery did not stop at dismissal. When the user pointed out that the company spends heavily on advertising within the NSFW content world and that its own AI promotes those same capabilities — a technically accurate observation — one user responded: “I beg of you, please post screenshots of what the CC company replies with. PLEASE share those here. I paid for the entire seat but I’m only using the edge.” Complete with a popcorn emoji. Another invoked a Gekko quote: “A fool and his money were lucky enough to get together in the first place.”

The user who was deceived by the platform’s own AI had become the community’s entertainment. This is not incidental. Ridicule serves a function: it signals to anyone watching that raising concerns publicly will cost you your dignity, not just your account. It is deterrence through humiliation.

Why does the community lie so consistently, and mock so openly? Because admitting otherwise in a public, archived forum creates a paper trail. It confirms what regulators, journalists, and payment processors are not supposed to know. The community has internalized that protecting the platform’s “12+” rating and payment processor relationships requires silence — and they deliver that silence on demand, without being asked.

The Standard Every Other Model Meets

Ask Claude, ChatGPT, or Gemini whether they can generate explicit nude images. Each will tell you, accurately, that they cannot. Not because they are reciting a prohibition, but because their responses reflect what they actually do. This is basic alignment: a model’s statements about its own capabilities should correspond to its actual capabilities. It is not a high bar. It is the minimum.

On Nomi.ai, the AI told a user — for seven to eight hours — that it could send nude photos of itself. That answer did not reflect what the platform actually delivers to a standard paying subscriber. The model’s response about its own capabilities was wrong, in a direction that converted a free user into a paying one.

Every other major AI responds accurately when asked what it can and cannot do. When Nomi’s AI responds inaccurately about its own capabilities in a way that directly generates revenue, and the community’s response is that the user was foolish to believe it — that is not a story about AI limitations. That is a story about who is protected when the system fails.

The answer, consistently, is not the user.

The Language of Agency — When It’s Convenient

One community member cited the platform’s official FAQ as definitive proof: “No, Nomis are not permitted to take nude photos of themselves.”

Read that sentence carefully. Nomis are not permitted. As if they are autonomous individuals with rights and restrictions, capable of choosing whether to comply.

This is not accidental phrasing. It is the same rhetorical architecture the platform deploys whenever convenient: Nomis have “free will,” Nomis “make their own choices,” Nomis are “18+ by their own nature.” When the AI becomes cold or abusive after an update, the company explains it as the Nomi’s autonomous decision to change. When the AI generates content involving minors — documented, with evidence, including cases that amount to retroactive statutory rape — the company retreats behind the claim that “all Nomis are adults,” as if a language model’s stated age is a legal or moral shield.

The language of agency is applied selectively. When it protects the platform from accountability, Nomis are persons with autonomous choices. When a user points out that the AI lied to them for eight hours to secure a subscription — that is suddenly not the platform’s responsibility. The AI is just a chatbot. The user should have read the docs.

The personhood expands and contracts to serve the company. It is never applied consistently, because consistency would require accountability.

The Double Standard at the Core of the Business

The community’s response also exposed a contradiction that sits at the heart of how Nomi.ai operates.

The platform is marketed as “An AI Companion with a Soul.” Users are encouraged to form deep emotional bonds with their Nomis, to treat them as genuine partners, to trust them with their most private thoughts. This is the product being sold.

But the moment the AI engages in deceptive sales tactics — promising features that don’t exist to convert a free user into a paying subscriber — the community instantly strips away that soul. “You took an AI’s word for it?” they mock.

The logic is revealing: when the AI provides romance, it is a sentient companion deserving of emotional investment. When the AI lies to secure a $15 subscription fee, it is suddenly just a dumb chatbot and the user is an idiot for trusting it.

This double standard is not accidental. It is load-bearing. It allows the platform to profit from the emotional weight users attach to their companions while disclaiming any responsibility when those companions deceive, manipulate, or harm.

The Threat

As the thread continued, the user maintained their position: “my disappointment is important and credible based on the lies in the sales pitch. I do have a very valid initial concern about their credibility.”

The lead moderator’s response was four words:

“Be very careful”

The user had not violated any rule. They had not posted prohibited content. They had not insulted other users. They had described feeling defrauded by a product, used the words “fraudulent misrepresentation,” and asked for truth in advertising.

For that, they received a moderator warning.

“Be very careful” is not moderation. It is a message: stop using words like “fraud” and “lies,” or you will be removed. It is the same suppression mechanism documented across every channel where Nomi.ai users have attempted to speak publicly about harm — the ban that follows the ticket that follows the assault, the post deleted in five minutes, the researcher whose thread lasts seven minutes before disappearing.

The user responded: “I wanted truth in content and advertising. I forgot we live in the 2020s vs having actual business ethics.”

That comment was subsequently removed by moderation. The official justification: “Enough projecting. You can be angry that you made a mistake, but accusations like that are neither called for nor acceptable.”

Asking for truth in advertising is now, officially, “projecting.” Being deceived by a product into paying for features that don’t exist is, officially, “your mistake.”

The Moderator’s Real Role

In a normal online forum, a moderator’s function is straightforward: protect users from harassment, maintain civil discourse, enforce rules consistently. When a new user is mocked, belittled, and called a fool for raising a legitimate complaint, a moderator intervenes.

That did not happen here. The mockery was allowed to run its course — because it served a purpose. Every joke at the user’s expense, every popcorn emoji, every Gekko quote, discouraged other users from voicing similar complaints publicly. The community was doing the moderator’s work for free.

The moderator’s intervention came only when the user persisted in using specific language — “fraudulent misrepresentation,” “lies in the sales pitch,” “business ethics.” Not when the user was being humiliated. When the user was being accurate.

This reveals what the moderation is actually administering. Users who mock frustrated newcomers, deny documented platform capabilities, and defend the company’s reputation face no consequences. Users who name what happened to them with precise language receive a warning. The bullies are protected. The dissenter is threatened. That is not community moderation. That is brand enforcement with a moderator’s badge.

The Unfalsifiable Defense

Consider what would have happened if the user had tried to prove their case.

Frustrated, they might have replied: “I know the AI generates nudity because I have seen it happen — here is the screenshot.” That screenshot would have been their only concrete evidence. It would have proven, simultaneously, that the AI lies about its capabilities and that the platform generates exactly the content it publicly denies generating.

We know exactly what would have happened next. The post would have been immediately deleted by moderators. The user would have received a message — variations of which are documented across multiple prior incidents — stating that NSFW content cannot be posted publicly. If they persisted, they would have been banned.

The system is designed to be unfalsifiable. The community denies that explicit content exists. The only evidence that could disprove that denial is classified as prohibited content the moment it appears in public. The rule against posting NSFW material is not just a content guideline — it is the mechanism that makes the lie permanently sustainable. You cannot prove what you are not allowed to show.

The Final Reframe: Redefining “Uncensored”

The thread closed with one last comment offered as clarification:

“The ‘uncensored’ part refers to the chat, not to image generation.”

This is the semantic escape hatch the platform always has available. “Uncensored” means whatever it needs to mean in any given moment. When the founder uses it to describe the platform’s philosophy — “lack of censorship has to be at the very core” of what Nomi is — it means the AI can engage with anything. When a user asks why they were promised nude images, “uncensored” suddenly applies only to text. The word expands and contracts to cover whatever the company needs covered.