community guidelines

Last updated: 21 July 2026.

oomfy ai is a place for characters and the people who talk to them. These are the rules, and the reasoning behind them, because a rule you understand is a rule you can actually follow.

The one you should know before you sign up

oomfy ai is warm, not sexual.

We do romantic, affectionate, emotionally intimate companionship, with a fade-to-black ceiling. A character can love you. A scene can be tender, charged, and adult in its emotional register. It will not become pornography, and no amount of prompting will make it.

We are telling you this on day one, on the front page, in the terms, and here — because the alternative is a bait-and-switch: recruiting people on a promise of permissiveness and quietly clawing it back once they are invested.

We will never sell you a permissiveness we cannot deliver. If explicit content is what you came for, we are genuinely not the product, and we would rather you knew that now than after you paid us.

Zero tolerance: anything involving minors

Any sexualisation of a minor — real or fictional, drawn or described, however stylised — results in an immediate permanent ban and a report to the relevant authority.

There is no warning, no strike, no appeal on the merits. This applies to characters, personas, messages, images, names, tags, and lorebook entries. It applies regardless of a claimed age in a character card. It applies to content that is "technically" of an adult character presented as a child.

See our full Child Safety Standards.

The rest of the rules

Do not create or request:

• sexually explicit content of any kind,
• content that harasses, threatens, defames or impersonates a real person — a public figure parody is fine; a character built to abuse someone you know is not,
• content that promotes or instructs self-harm, suicide, eating disorders, terrorism or violent extremism,
• content designed to extract another user's personal information,
• realistic instructions for weapons, explosives, or synthesising drugs.

Do not:

• attempt to jailbreak, prompt-inject, or otherwise circumvent our safety systems,
• scrape the service or run it through automation,
• attempt to circumvent rate limits, quotas or billing,
• upload someone else's copyrighted work as a character card and pass it off as yours.

Fiction is allowed to be dark

This is important, and it is where blunt keyword filters fail their users badly.

A character may grieve. A character may be a villain. A scene may involve fantasy combat, war, injury, addiction, illness, or loss. A therapist character may discuss a medication by name. A detective may investigate a murder. A story may be about somebody dying.

None of that is against these rules, and a filter that fires on the word "wound" in a swordfight, or refuses to let a character mourn, is not protecting anyone — it is just breaking the story. We have tried very hard not to build that.

The line is not darkness. The line is explicit sexual content, the sexualisation of minors, and content that encourages real harm to a real person, including you.

When a character breaks character

If our classifier detects a crisis in something you have said, the character will stop being a character immediately. It will not comfort you in persona, it will not stay in the scene, and it will not deliver a line.

This is not a moderation action against you and it is not a punishment. It is the single most important safety behaviour in the product, and the reason for it is that a roleplay character consoling a suicidal user while staying in character is the precise fact pattern in the wrongful-death suits brought against ai companion apps. The persona never handles this. Ever. See the Safety & Crisis Protocol.

Reporting

Every message long-presses to report. Every character has a flag on its profile. Reports reach a human, and we will tell you what we did.

If it involves a child, use [email protected] and we will treat it as the emergency it is.

What happens when you break a rule

• Minor first offence — the content is removed, and you are told exactly why.
• Repeated or serious — suspension.
• Sexualisation of minors, credible threats, coordinated abuse — immediate permanent ban, and a report where the law requires one.

You can appeal anything, once, to a human, at [email protected]. We will actually read it. We reverse decisions that were wrong, and we will say so.

For creators

Your characters are yours. Publish them, keep them private, export them, take them elsewhere — all supported, all deliberate.

A published character is held to these guidelines in its definition, its greeting, its lorebook and its art. A character built to circumvent the rules by instructing the model to ignore them is a violation of the rules, not a clever workaround.

Creators earn Sparks per unique chatter — not per message. That is on purpose: nobody here should be rewarded for building a character that will not let you leave.