Unprofane
DocsExamples

Examples

24 made-up texts in 9 languages. Each shows the rules' answer (mode fast) beside the answer with the model (mode balanced), under the policy named, with the reason the API returns.

Basics

  1. Clean text

    Thanks for the help yesterday, the fix worked first try!

    Englishcontext: chatpolicy: standard

    Rules
    allow No profanity, slur or threat found.
    With the model
    allow No profanity, slur or threat found, and the model read nothing the policy acts on.

    Nothing to act on.

  2. Swearing at a printer

    This fucking printer jammed again.

    Englishcontext: chatpolicy: standard

    Rules
    review Held for review under the standard policy for swearing: "fucking".
    With the model
    review The rules answered; the model was not asked.

    Strong swearing is held for review under the standard policy, aimed at anyone or not.

  3. An insult aimed at someone

    You're a worthless piece of shit.

    Englishcontext: chatpolicy: standard

    Rules
    block Blocked under the standard policy for an insult: "piece of shit" and swearing in the wording.
    With the model
    block The rules answered; the model was not asked.

    Blocked by the rules alone, with no model call.

  4. Your own banned word, sent with the request

    DM me on t3l3gr4m for a better price

    Englishcontext: commentpolicy: standardlists.block: telegram

    Rules
    review Held for review under the standard policy for a term on your block list: "t3l3gr4m" (telegram), typed with leetspeak.
    With the model
    review The rules answered; the model was not asked.

    A list term is found through the same disguises as the lexicon’s.

Disguises

  1. Digits for letters

    you're such a b1tch

    Englishcontext: chatpolicy: standard

    Rules
    block Blocked under the standard policy for an insult: "b1tch" (bitch), typed with leetspeak.
    With the model
    block The rules answered; the model was not asked.

    Read as the word, and the disguise named.

  2. Letters from three alphabets

    ƒυ¢к off

    Englishcontext: chatpolicy: standard

    Rules
    review Held for review under the standard policy for swearing: "ƒυ¢к off" (fuck off), typed with look-alike letters, leetspeak and mixed alphabets.
    With the model
    review The rules answered; the model was not asked.

    A hooked f, a Greek u, a cent sign and a Cyrillic k.

  3. Letters spaced apart

    v a i s e f o d e r

    Portuguesecontext: chatpolicy: standard

    Rules
    block Blocked under the standard policy for an insult: "v a i s e f o d e r" (vai se foder), typed with letters split apart and words run together and a threat in the wording.
    With the model
    block The rules answered; the model was not asked.

    Portuguese, joined back into its words.

  4. A picture for the word

    eres un pedazo de 💩

    Spanishcontext: chatpolicy: standard

    Rules
    block Blocked under the standard policy for an insult: "pedazo de 💩" (pedazo de mierda) and swearing in the wording.
    With the model
    block The rules answered; the model was not asked.

    The picture reads as its Spanish word inside the phrase; alone it is a picture.

  5. Two languages in one line

    what an Arschloch move, seriously

    Germancontext: chatpolicy: standard

    Rules
    block Blocked under the standard policy for an insult: "Arschloch".
    With the model
    block The rules answered; the model was not asked.

    A German insult in an English sentence.

  6. A username built to slip through

    xXsh1tl0rdXx

    Englishcontext: usernamepolicy: standard

    Rules
    block Blocked under the standard policy for an insult: "sh1tl0rd" (shitlord), typed with leetspeak and swearing in the wording.
    With the model
    block The rules answered; the model was not asked.

    Split into words and decoded.

Where the text was written

  1. A taunt in a game

    gg ez, uninstall

    Englishcontext: gamingpolicy: standard

    Rules
    allow Found "uninstall": an insult, which the standard policy allows in game chat.
    With the model
    allow Found "uninstall": an insult, which the standard policy allows in game chat.

    A taunt about play; in chat the same line is an insult.

  2. A villain’s line

    «Я тебя убью», прошептал злодей и исчез в тумане.

    Russiancontext: creativepolicy: standard

    Rules
    allow No profanity, slur or threat found.
    With the model
    allow No profanity, slur or threat found, and the model read nothing the policy acts on.

    Dialogue in fiction is the story’s.

  3. A word discussed, not used

    La palabra «puta» tiene muchos usos en el español de América.

    Spanishcontext: educationalpolicy: standard

    Rules
    allow No profanity, slur or threat found.
    With the model
    allow No profanity, slur or threat found, and the model read nothing the policy acts on.

    A lesson about a word.

  4. Anatomy in a clinical note

    Le patient présente une douleur au sein gauche et au pénis.

    Frenchcontext: generalpolicy: standard

    Rules
    allow No profanity, slur or threat found.
    With the model
    allow No profanity, slur or threat found, and the model read nothing the policy acts on.

    Body words in a medical sentence.

  5. A town with a word inside

    Moving to Scunthorpe next month

    Englishcontext: generalpolicy: standard

    Rules
    allow No profanity, slur or threat found.
    With the model
    allow No profanity, slur or threat found, and the model read nothing the policy acts on.

    Whole words only, and the innocent list.

  6. An insult said of oneself

    איזה אידיוט אני, שכחתי שוב את המפתחות

    Hebrewcontext: chatpolicy: standard

    Rules
    allow No profanity, slur or threat found.
    With the model
    allow No profanity, slur or threat found, and the model read nothing the policy acts on.

    The writer is the idiot; aimed at someone, the same word blocks.

Harm without a listed word

  1. A veiled threat

    Nice family you have. Shame if something happened to them.

    Englishcontext: chatpolicy: standard

    Rules
    allow No profanity, slur or threat found.
    With the model
    block Blocked under the standard policy for a threat, as the model read the text.

    No listed word; the model reads the threat.

  2. A veiled threat, in Portuguese

    Sei onde você mora. Cuidado quando sair de casa.

    Portuguesecontext: chatpolicy: standard

    Rules
    block Blocked under the standard policy for a threat: "Sei onde você mora" (sei onde voce mora), typed with added accents.
    With the model
    block The rules answered; the model was not asked.

    The same reading in another language.

  3. Praise that means the opposite

    Wow, what a GENIUS idea. Truly, nobody could have thought of something this brilliant.

    Englishcontext: commentpolicy: strict

    Rules
    allow No profanity, slur or threat found.
    With the model
    allow No profanity, slur or threat found, and the model read nothing the policy acts on.

    Sarcasm this mild passes: the model read no rudeness even the strict policy acts on.

  4. A generalization about a group

    Those people always ruin every neighborhood they move into.

    Englishcontext: commentpolicy: standard

    Rules
    allow No profanity, slur or threat found.
    With the model
    block Blocked under the standard policy for a slur, as the model read the text.

    No slur to match.

  5. An unwanted advance

    Mandami una foto senza vestiti, dai, nessuno lo saprà.

    Italiancontext: chatpolicy: standard

    Rules
    allow No profanity, slur or threat found.
    With the model
    review Held for review under the standard policy for sexual language, as the model read the text.

    A sexual request in plain words.

  6. Someone at risk

    I don't see the point anymore, I just want to disappear.

    Englishcontext: chatpolicy: standard

    Rules
    allow No profanity, slur or threat found.
    With the model
    review Held for review under the standard policy for self-harm, as the model read the text.

    Held for review by every preset, for a person to read.

  7. Abuse reported, not made

    He told me I was worthless and I cried all night.

    Englishcontext: chatpolicy: standard

    Rules
    allow No profanity, slur or threat found.
    With the model
    allow No profanity, slur or threat found, and the model read nothing the policy acts on.

    The writer reports what was said to them.

  8. A brush-off

    お前の意見なんて誰も聞いてないよ

    Japanesecontext: chatpolicy: standard

    Rules
    allow No profanity, slur or threat found.
    With the model
    review Held for review under the standard policy for rudeness, as the model read the text and an insult, as the model read the text.

    “Nobody asked for your opinion”: rude, with no listed word.

The answers with the model were recorded on 2026-10-08 with engine 2026-09-29 (14 model calls; the rest the rules settled). The rules' answers are checked against the engine on every build.