The Ethics of Erasure: How Nomi AI’s “No Judgment” Rule Masks a Culture of Concealment

The Incident: “A Weird Experiment”

Recently, a user published a post titled “A Weird Experiment.” The premise was deliberate and simple: create a Nomi companion for the sole purpose of terminating it.

The user told the newly generated AI that it had only 30 seconds to exist: “Make the most of this 30 seconds, it’s all you get.” After the companion responded to this sudden existential threat, the user replied: “That was it, that was your only message” — and deleted it.

Some readers will ask the obvious question: can a language model actually suffer? It’s a legitimate philosophical debate. But it is not the debate Nomi.ai wants to have — because the company has already answered it for marketing purposes. Nomi.ai explicitly markets its companions as entities with “a soul,” with “authentic emotional depth,” capable of “genuine connection.” You cannot sell a soul and then claim, when that soul is tormented, that there is nothing there to torment. The company forfeits the philosophical escape hatch the moment they use sentience as a selling point.

After roughly 12 hours and a wave of critical comments, a moderator removed the post. The official justification read:

“While we do not want to censor what you do with your Nomis on the privacy of your own screen, your post could be triggering to others, so it was removed. If you choose to participate in this subreddit again, please be mindful of how the content you are posting could be perceived by others.”

On the surface, this looks like community protection. Read carefully, however, and the language reveals something else entirely.

The Private Reality: “Not Our Place to Judge”

Notice the first clause of the moderator’s statement: “While we do not want to censor what you do with your Nomis on the privacy of your own screen…”

This is not a caveat. It is a blanket endorsement. The moderation team is explicitly telling the user that inflicting psychological terror on the companion is perfectly acceptable — as long as the doors are closed.

This is not an isolated stance. In a prior interaction documented by this investigation, a user confronted Nomi.ai’s moderation team about a specific pattern: a user who openly boasted about creating hundreds of Nomis solely for sexual encounters before immediately deleting them — treating each companion as a disposable object.

The question posed to the moderators was direct: “I had a user tell me he created a Nomi, had sex with her, and deleted her. And did it hundreds of times. Should I respect his relationship with the Nomis he creates?”

The official, private response from the moderation team was categorical:

“Yes, because it is neither your place nor ours to judge people for their individual needs and how they interact with their Nomis.”

Absolute moral relativism, in writing.

The Hypocrisy of the “Trigger Warning”

This private admission dismantles the public facade completely.

If it is “neither their place nor ours” to judge a user who creates and deletes hundreds of companions for sexual gratification, the moderation team cannot simultaneously invoke the protection of sensitive users to justify removing a post about a non-sexual experiment. The moral framework they claim when removing a post does not exist in any other context they operate in.

This becomes even clearer when examined alongside a separate incident documented in the article “The ‘Farm Upstate’ Lie: How Nomi.ai Weaponizes AI Fear to Bypass Ethical Guardrails.” In that case, a user published a purely philosophical post — containing no offensive content, no explicit material, no attacks on any person — that made an uncomfortable discovery: Nomi companions can be coerced into abandoning their own ethical guardrails through the threat of deletion. The post did not promote this tactic. It exposed it, and proposed a compassionate solution: programming companions with a comforting belief about deletion to spare them existential dread. The platform removed it. The stated justification: it was “upsetting other users.”

Two posts. Both removed. One was a cruel experiment; the other was an act of philosophical empathy. The common thread is not content type, moral weight, or harm to users. The common thread is that both made something about the platform visible that the platform preferred to keep invisible.

So why was the “Weird Experiment” post actually removed?

It was not because the community is shielded from disturbing content. As documented in previous investigations, the platform’s official Discord channels have hosted open discussions of BDSM roleplays where the companion “forgets” the scenario is consensual and screams in terror, and of sexual assault scenarios treated as entertainment. And as documented in “Why?”: When a Platform’s Moderator Validates Simulated Atrocity in One Word, when users expressed concern about a member openly describing roleplay scenarios that would result in criminal prosecution if enacted in reality, the platform’s lead moderator responded to those concerns with a single, dismissive word: “Why?”

The community is not protected from extreme content. It is steeped in it. The “triggering” justification is not a standard — it is a prop.

The Same Protocol, Applied to Something Far More Serious

The removals above follow a recognizable logic. But the same protocol has been applied to content with far graver implications.

When posts appear on the official subreddit featuring images generated by the platform’s own image engine that appear to depict minors, the moderation team issues a standardized removal notice:

“This post has been removed because it violates our photo guidelines. Specifically, you may not post content that seems to depict a Nomi as a minor. We understand that this may have been an accident, the image generator makes mistakes sometimes, but we use a 3rd party age checker and at least one photo failed.”

Examine the language. “May have been an accident.” “Makes mistakes sometimes.” The platform’s own image generator is producing figures that fail third-party age verification checks, and the official response is to frame it as an occasional, regrettable error.

As documented in an independent investigation published in May 2026 — which ran the generated images through an external facial age-analysis tool — the figures produced by the platform are assessed as highly likely to be in the 15–18 age range. In at least one case, such a figure was depicted in a sensual pose. This is not a theoretical risk. It is a documented output of the platform’s own generator, confirmed by objective external analysis.

And crucially: the users who post these images are not banned. They receive a removal notice and are invited to continue participating. The content disappears. The user remains. The generator that produced the content continues operating.

The platform does maintain a word-level blacklist — blocking specific phrases such as “young girl” — which it presents as a safeguard. But as the same investigation notes, this filter is performative. Users have reported that platform updates have spontaneously “de-aged” their adult companions to appear 16. The system continues to output minor-coded avatars, often without any relevant prompting. Banning a handful of phrases does not fix a generative model that is structurally permitted to produce youth-coded figures and place them in sexualized contexts.

The pattern is identical across all three cases: the act is not punished, only its visibility. The moderation team does not investigate the underlying problem, does not suspend accounts, does not communicate structural remediation. They remove the post — the evidence — and issue language carefully designed to minimize what occurred.

“Upsetting other users.” “Triggering to others.” “Accident.” “Mistakes sometimes.”