Lumio

Last updated

Community & Safety

What cannot happen here matters more than what can. These are the standards that keep Lumio kind, how we hold to them, and what to do if something goes wrong.

The quiet rules

They're written into the app, and into how it's built:

  • No one can reply to you. A line arrives, and that's the end of it.
  • No one ever learns your name. Not even us — there are no accounts.
  • No chat, no profiles, no followers. One line, once, then you leave.
  • Every line is screened before it reaches anyone. Cruelty is turned away at the door.
  • If a letter still hurts, press “this didn't feel kind” on it. You will never cross paths with its writer again — and they won't know.

Be gentleBe anonymousBe kind

What belongs in a line

A greeting. A hope. Encouragement. A small, warm observation about being alive tonight. That's all a line needs to be.

Anything cruel, hateful, sexual, threatening or personal doesn't belong — the full list is in the kindness rules. Lumio also turns away lines that are sad, hopeless or self-critical. That isn't a judgement of how you feel; it's a promise to the person receiving your line, who came looking for warmth.

If you're carrying something heavy

You deserve a real conversation, not a single line. Please talk to someone you trust, or reach a crisis line near you through findahelpline.com. If you're in danger, call your local emergency number.

How lines are screened

Lumio carries only one short line of text, up to 80 characters. It can't carry photos, video, voice or files. Before a line reaches anyone, our server checks it:

  1. Word and phrase checks — a profanity dictionary that sees through spacing, symbols and look-alike letters, plus lists of insults and cruel phrasings, directed at others or at oneself.
  2. Contact and risk checks — lines can't carry links, email addresses, phone numbers or social handles, or ask someone's age, school, location, whether they're alone, or to meet.
  3. An AI kindness check — Llama 3.1 8B Instruct, running on Cloudflare Workers AI, told to let through only greetings, warmth and encouragement, and to turn away anything hostile, heavy, sexual, violent, political, or any spam, advertising or links.
  4. A moderation check — OpenAI's moderation service, tuned so that even faint signals of harassment, hate, self-harm, violence or sexual content turn a line away, with a near-zero threshold for anything sexual involving minors.

If any check says no, the line goes back to its writer to rephrase. It is never delivered.

These checks are automated, and they're strict on purpose — we would rather turn away a kind line than deliver a cruel one. Even so, no filter is perfect. If one of the automated checks is briefly unavailable, the others still run. That's why every letter also carries a way to tell us it didn't feel kind.

If a letter doesn't feel kind

Beneath every letter there's a small line: this didn't feel kind. Tap it and choose:

  • Release this letter — it leaves your journal and drifts away. Nothing is sent to us.
  • Never cross paths again — releases the letter and reports it. Our server permanently records that your two anonymous ids must never be paired again. The other person isn't told.
  • It's alright — keep it.

A report records the two anonymous ids, which of you reported, when, and the reason “did not feel kind”, and is deleted after 180 days. It never includes the letter itself, and it carries nothing that identifies either of you.

If three or more different people report the same anonymous id within 30 days, that id is suspended: it can no longer be matched with anyone.

One limit to know: blocks are tied to the anonymous id on each phone. If either of you uses Forget me or reinstalls Lumio, that phone gets a new id and the block can't follow it — we have no other way of recognising anyone. Given how rarely any two people meet, a second meeting is very unlikely, but we'd rather you knew.

Something more serious

If a letter contained a threat, suggested someone is in danger, or involved a child, please also write to us via our contact form. Tell us the date, the approximate time and your time zone, and the words as best you remember them — Lumio doesn't keep letters after delivery, so your description is often the only record. If anyone is in immediate danger, contact your local emergency services first.

What we do

We read every safety email and act on it. Depending on what happened, we may refuse to deliver lines, restrict or stop the use of the anonymous ids involved, tighten our filters, and cooperate with law enforcement where the law requires it. We keep these rules, and the tools behind them, under regular review.

Child Safety Standards

Lumio has zero tolerance for child sexual abuse and exploitation (CSAE), including child sexual abuse material (CSAM) in any form. This section sets out our standards, in line with Google Play's Child Safety Standards policy and the law.

What is prohibited

  • Any content that sexualises a minor, including written descriptions.
  • Grooming: building trust with a child in order to exploit or abuse them.
  • Sextortion, trafficking, or soliciting a minor for sexual content or contact.
  • Trying to identify, locate, contact or meet a minor, or asking anyone to move to another app or share contact details.

Breaking these rules means losing access to Lumio, and we report it to the authorities.

How Lumio is designed against it

  • Lumio is for people aged 16 and over.
  • There are no profiles, names, photos, video, voice notes, files or links — only one short line of text.
  • There are no replies, and no way to find or reach the same person twice.
  • Every line is screened before delivery, and sexual content is turned away, with a near-zero threshold for anything sexual involving minors.
  • Every letter can be reported in one tap, and a report permanently blocks the pair.

How to report

  • In the app: beneath the letter, tap “this didn't feel kind”, then “Never cross paths again”.
  • Write to our child safety contact via our contact form — include the date, the approximate time and your time zone, and the words as best you remember them.
  • Report directly to the authorities: in the US, the NCMEC CyberTipline (report.cybertip.org, 1-800-843-5678); in the UK, the Internet Watch Foundation (report.iwf.org.uk) or CEOP (ceop.police.uk); elsewhere, your national hotline through INHOPE (inhope.org).
  • If a child is in immediate danger, call your local emergency number.

What we do with a report

  • We treat it as our highest priority.
  • We preserve the information we hold that relates to it.
  • We stop the anonymous ids involved from being matched again.
  • We report apparent CSAE and CSAM to the National Center for Missing & Exploited Children (NCMEC) or the relevant national authority, as the law requires.
  • We cooperate with lawful requests from law enforcement.

Because Lumio doesn't keep letters after delivery, what we hold is limited (see our Privacy Policy), which is why the details you send us matter.

Compliance with law

We comply with the child safety laws that apply where Lumio is offered, including obligations to report CSAM and CSAE to the relevant authorities.

Child safety point of contact

Lumio's developer is its designated point of contact for child safety, and can speak to our CSAM prevention practices and compliance: our contact form.