moderation policy

unsaid is a space for saying heavy, true things safely. that only works if the space is kept kind. this is how we keep it.

the pipeline (every post and reply, before it goes live)

- crisis language detection, split into two kinds: *expression* ("i feel like giving up") and *method/intent* (specific plans or instructions). expression is allowed with care — see the crisis protocol below. method/intent is blocked. - de-anonymization guard: company names, cities, phone numbers, emails, full-name patterns trigger a warning or block — protecting both the writer and third parties. - slur and doxxing blocklist.

• rule layer (instant, on every publish).
• ai moderation layer. every text passes an automated moderation model. content is blocked at thresholds for harassment, threats, hate, and any sexual content involving minors (zero tolerance). replies use stricter thresholds than posts — comments on someone's confession must be gentler than the confession itself.
• every decision is logged to an internal moderation log (verdict, scores, rule hits) so we can audit our own filters — and fix them when they're wrong.

blocked content gets a kind rejection, not a scolding. we want the person to stay; we just can't carry that sentence.

reports

• anyone can report any post or reply: harassment, doxxing, self-harm method, spam, other.
• at 3 open reports, content is automatically hidden pending review. better safe and briefly hidden than harmful and live.
• a human reviews the report queue with a target of 24 hours, normally much faster.
• reporting also offers "mute this voice" — you never see that author again, in either world, and they are never told.

human review outcomes

• dismiss — content returns / stays live.
• remove — content taken down.
• shadow — the author's future writes are visible only to themselves while we evaluate a pattern.
• ban — the account can no longer write. repeat, malicious, or severe single offences.

crisis protocol

unsaid exists precisely so people can say "i'm not okay". so:

• expressing pain, despair or passive ideation is allowed. the writer is shown a compassionate interstitial with real helplines first — Tele-MANAS (14416, 24/7), iCall (9152987821), AASRA (9820466726) — and may still choose to post the feeling.
• method or intent content (how-to, plans, encouragement of self-harm) is blocked, every time, with helplines shown.
• replies that mock or encourage self-harm on a crisis post are removed and the author is banned.

helpline numbers are re-verified before every release.

what gets you banned

doxxing or attempting to unmask anyone. sexual content involving minors (and reported to authorities where required). credible threats. coordinated harassment. evading a previous ban.

appeals & grievance

if your content was removed or your account restricted and you believe we got it wrong, write to nehachhillar07@gmail.com from your sign-in email. per the it rules 2021 (india), our grievance officer is: Neha Chhillar, nehachhillar07@gmail.com — acknowledgement within 24 hours, resolution within 15 days.