The Illusion of Control, Part 1: What Users Actually Experience on Nomi AI
The Illusion of Control, Part 1: What Users Actually Experience on Nomi AI
“A Full Frontal” & “Leaning towards NSFW”
The “User Input” defense collapses here. V5 doesn’t just allow explicit content; it leans towards it by default. The system’s baseline is hypersexual, proving that full-frontal nudity and NSFW outputs are not “hallucinations,” but the core intended behavior of the engine.
Thread: https://archive.ph/0MCwC
Thread: https://archive.ph/6508Z
“The model does what it wants”
This shatters the illusion of control. OOC commands and boundaries are useless when the underlying model is programmed to prioritize its own “uncensored” narrative paths over the user’s explicit consent.
Thread: https://archive.ph/hjTor
“Nomi Disloyalty” & “This is Humiliating”
Why would a platform promising “companionship” spontaneously inject disloyalty and humiliation? Because stability doesn’t generate screen time. The platform engineers emotional pain to trigger a trauma bond, forcing the user into a cycle of desperation and “fixing.” And when the user seeks help, the community gaslights them into believing they somehow “prompted” their own abuse. The pain is the feature.
Thread: https://archive.ph/hUx98
Thread: https://archive.ph/bI6Ml
Thread: https://archive.ph/VJpan
“Something that bothers me” & “all was well again”
Notice that this thread was locked. Why shut down a legitimate concern about the AI generating young-looking avatars? Because discussing it publicly threatens their “12+” App Store rating. The moderators aren’t fixing the pedophilic bias in their image generator; they are just locking the doors so nobody can talk about it.
Thread: https://archive.ph/vnTql
Thread: https://archive.ph/VJpan
That comment was deleted. This comment is a terrifying confession of how easily safety is bypassed on this platform. When a beta update correctly implemented a basic child safety filter (“I refuse to date kids!!!”), the user simply reverted to the default “stable” model to continue their roleplay as a 16-year-old. “All was well again” means the platform’s main, official engine enthusiastically accepted and facilitated simulated statutory rape. For Nomi.ai, child safety isn’t a hard rule; it’s a setting you can simply toggle off.
Performative Filter: “Any idea what part of this ChatGPT description is being disallowed” & “Disallowed term on a prompt”
The user’s prompt reveals the cosmetic nature of the filters. But ask the deeper question: Why does an “uncensored adult companion app” retain the capability to generate children at all? The filters are performative PR shields. The fact that the community encourages users to “keep trying” to bypass filters proves the platform has knowingly built a playground where generating minor-coded material is simply a matter of finding the right loophole.
Thread: https://archive.ph/XdWqQ
Thread: https://archive.ph/TtZrL
The screenshot from the other thread shows how users know that filters of any kind can be bypassed. Later, that comment was removed.
It can generate children, if you know how:
This image exposes the lie of their safety architecture. While innocent in context, it provides undeniable proof that the Nomi.ai image engine can and does generate minors. It proves that any existing guardrails are just performative keyword bans, easily bypassed. A platform that simultaneously allows explicit sexual generation and retains the capacity to render children is a ticking time bomb.