# Twelve things, checked before a message sends.

> Exactly what Naven screens for, exactly what it deliberately leaves alone, how age bands work, and what is stored. Written so a school, a library, or a parent can assess it against their own internet safety policy without reading any code.

- Page: https://navenapp.com/safety
- Last updated: 2026-10-19
- Published by: Naven, a free messaging app that runs in a web browser (https://navenapp.com/about)

- **Screening:** Every message, before it sends
- **Age bands:** Under 13 · Teen · Adult
- **Under-13 messaging:** After a parent's verified consent, with the strictest screening
- **Member directory:** None — nobody is browsable

**This page describes software that tries; it promises nothing.** Screening is automated, best effort, and misses things — including obvious ones. Screening, age checks, consent checks and parent controls are algorithms that make mistakes in both directions and can, and sometimes will, fail; everything below says what they are designed to do and may do, not what they will do. Naven offers no expectation of privacy and no expectation of security, is not a child-safety product, and is used at your own risk. The [Terms of Service](https://navenapp.com/terms) govern, in particular sections [1](https://navenapp.com/terms#privacy), [2](https://navenapp.com/terms#security), [11](https://navenapp.com/terms#promises) and [12](https://navenapp.com/terms#liability).

Naven is a social platform for messaging, voice and video calling, and communities, used by adults, teenagers and children. Every message is checked **before it is delivered**, against a policy of exactly twelve categories:

**Spam and scams · malware and malicious links · harassment and bullying · hate speech · credible threats of violence · sexual exploitation or abuse · illegal transactions · doxxing and privacy violations · encouraging self-harm · graphic violence · sexual content · drugs.**

**Display names are screened too, and to a stricter standard than messages.** A name is shown beside everything an account ever sends, to people who never chose to hear from it — including on a lock screen. So a name may not carry a slur, pose as Naven staff, or be a link or a phone number. Naven screens a name when it is saved _and_ again every time it renders one, so a name that fails reaches nobody.

**Self-harm is the one category Naven does not block.** Somebody telling a friend they are struggling is doing the right thing, and a product that deletes that message teaches them not to say it again. Naven delivers it, and offers a helpline alongside it. What _is_ refused is content encouraging or instructing somebody else to hurt themselves, which is a different act entirely.

**What happens to an account.** A refused message is not delivered, and for almost everybody that is the whole consequence. Sending is paused only for **serious conduct that keeps happening** — refusals for threats, bullying, hate speech or slurs, scams, malware, doxxing, illegal sales, encouraging self-harm or graphic violence on **three different days within 30**. Then an under-13 account waits for a parent or guardian, and anybody else is suspended for one to four weeks depending on the most serious thing in the pattern. The one exception: sexual solicitation, or anything sexual involving a minor, suspends on the first instance for five weeks. An account **blocked twice** may be asked to confirm its age before it keeps using Naven Social and the Forum — an adult account with AgeVerif, an account under 18 by linking a parent; one already linked to a parent carries on, and the parent is told. The full scale is in [the Terms](https://navenapp.com/terms#enforcement).

**Slurs are screened for every member, whoever they are aimed at.** **Swearing comes in three tiers.** Mild words — “hell”, “damn”, “crap”, “hate”, “sucks” — are delivered for everybody, under-13s included. Abbreviations such as “wtf” and “lmfao” are delivered for 13 and over and refused for under-13s. Strong swear words, spelled out or disguised, are refused for everybody — but a refusal for swearing is never a warning, is never emailed, and never counts against the account. Whoever wrote a refused message is told which rule it met rather than being left to guess. This page is updated whenever that changes, because a safety claim that no longer matches the code is worse than no claim at all.

**Contact details.** An email address or a phone number can be shared. A home address asks the sender to confirm it is theirs, or that they have permission to share it; that confirmation is recorded and the message is read again later. Accounts for somebody under 13 cannot share an address or a phone number at all. Posting somebody else's address to expose or target them is doxxing and is refused.

**What is still not screened matters as much as what is.** Naven does _not_ screen insults, disagreements, dark humour, or someone saying they are struggling. "You are being an idiot" and "I hate this" are things people say to each other, and an app that reads a private message and refuses to deliver an argument is not protecting anyone. Naven does not claim alignment with the Children's Internet Protection Act.

## 1 Pre-send screening On for every message

Every message is checked against the twelve categories before it is delivered — not reported afterwards. The check runs in layers, so it never depends on one provider being available, and no message is refused on one model's say-so alone: every refusal is put to a second model that reads Naven's written policy, in the message's own language, before it stands.

- **Deterministic pattern screening** Runs on every message. No network call, no third-party quota, and it cannot be skipped or switched off by anyone. Covers child sexual abuse material, self-harm encouragement, credible threats, sexual content, hate speech and slurs, drug sales, and known grooming phrasing. Text is normalised first, so spacing, punctuation and leetspeak do not evade it.
- **Classifier screening** A hosted moderation model reviews anything the first layer allows through.
- **Fallback classifier** A second, independent provider takes over whenever the first is unavailable or out of quota.
- **A second reading of every refusal** When any layer above would refuse a message, a larger model is given Naven's full written policy — what each category means, what is explicitly allowed, and how to read messages written in other languages or mixed ones — and decides whether the refusal stands. It exists because the layers above, read together, were refusing ordinary messages in Spanish, Polish, Urdu and Dutch. Slurs, plainly spelled swear words, and an under-13 account's own address or phone number are not second-guessed, and for an under-13 account the rules' refusal stands.
- **A review within hours, in context** Every refused message is read again within a few hours, this time alongside the messages before it in the same conversation. If the refusal was a mistake, the warning is withdrawn, any pause or suspension it alone caused is lifted, and the member is emailed to say so. Anything the review is unsure about is left for a person to look at.

### How the layers combine

Layers two and three can only _add_ blocks, never remove them. If both are unreachable, the verdict falls back to layer one rather than allowing unscreened content through — the system fails closed, not open. Naven AI is screened on both sides: the question before it reaches a model, and the answer before it reaches the member.

## 2 Age bands and protection levels

Every account is asked for a birth year once at sign-up. Naven derives a coarse band from it and **stores only the band** — the birth year itself is never written to the database.

| Band | Age | Protection applied |
| --- | --- | --- |
| Under 13 | 0–12 | Strictest filtering, plus blocking of addresses, phone numbers and ID numbers written in messages. Cannot send, post, upload, call or use Naven AI until a parent's signed consent is verified under COPPA; cannot share a location; also paused pending guardian approval if serious refusals happen on three different days within 30. The parent can turn on reading the words of a message the screen stopped. See the [children's privacy notice](https://navenapp.com/privacy/children). |
| Teen | 13–17 | Sexual content, drug-sale and grooming filters applied in addition to the universal categories. A parent or guardian agrees to Naven's terms for them — on the member's device or from an emailed link; after seven days, sending waits until they have. A linked parent can read the words of a message the screen stopped (from 19 October 2026, kept 30 days). Blocked twice: links a parent to carry on. See [the terms](https://navenapp.com/terms#parents). |
| Adult | 18+ | Universal categories only: CSAM, credible threats, hate speech, harassment and self-harm encouragement. Blocked twice: may be asked to confirm 18 or over with AgeVerif before using Naven Social and the Forum again. |

A member can correct their band, but can never move themselves into a less protected one. That restriction lives in the database security rules, not in the browser, so it holds even against a modified client.

## 3 Parental consent for under-13 accounts Before the first message

An under-13 account **cannot send anything until a parent's consent is verified**, as the US Children's Online Privacy Protection Act requires. The parent is emailed a notice, confirms they are an adult with AgeVerif, signs a one-page consent form and uploads a photo of it, and Naven's software checks it; a school may consent instead for accounts it creates for its own students. If consent does not arrive within 30 days, the account and everything about it is deleted. Naven AI needs the parent's separate permission. See the [children's privacy notice](https://navenapp.com/privacy/children).

A second gate fires on **conduct**. An under-13 account whose serious refusals keep happening — threats, bullying, hate, scams and the like, on three different days within 30; never swearing — has sending paused until a guardian confirms it, and a guardian is the _only_ way an under-13 account can clear it: the code-to-your-own-inbox route adults get is not offered. That verification record can only be written by Naven's server, and the security rules reject any attempt to write it from a browser, so a child cannot grant themselves consent by editing anything on their own device.

## 4 Monitoring of minors' activity

Where an institution has its own obligation to monitor minors' activity, Naven supports it by screening every message at the point of sending and recording safety events on the server, under records the member can neither read nor clear. Naven does not read private conversations for any purpose other than this screening, and does not sell or share message content.

**What a parent sees.** A parent or guardian linked to an account under 18 sees who it talks to, when, how much time it spends on Naven and which categories the screen stopped. From 19 October 2026 they can also read the words of a message the screen stopped it sending, for 30 days — on by default for 13 to 17, and under 13 only once the parent turns it on. A parent never sees a message that was delivered, anything anybody else wrote, or a stopped picture or voice note (only that one was stopped). The member is told in the app. Like everything on this page, this is automated and can be wrong or fail.

## 5 No public directory, no stranger discovery

Naven publishes no browsable member directory and has no algorithmic feed that surfaces strangers. A conversation can only begin when someone already knows the other person's email address, and that address is looked up as a one-way hash, so no list of members can be enumerated from it. Email addresses are never shown to other members, and being named in response to such a lookup is opt-in.

## 6 What Naven stores

#### ✓ Stored

- Display name and optional profile photo
- Age band (not the birth year)
- Theme preference
- Messages and their delivery state
- Conversation membership
- For a linked parent: the words of a stopped message, 30 days

#### ✕ Not stored

- Birth year
- Precise location
- Contact lists
- Browsing history
- Advertising identifiers

#### ✕ Not done at all

- No advertising
- No sale of personal information
- No profiling of minors
- No engagement ranking

## 7 Reporting and enforcement

Blocked messages are refused at the point of sending, with an explanation to the sender. Members can mute a conversation, and messages can be unsent by their author within two minutes.

To report content or an account, or to ask a question about this policy:

## 8 Known limits Read this

Naven states these plainly rather than implying more coverage than exists. An institution assessing Naven should assume exactly this much and no more.

- **Picture and voice-message checking is limited and unreliable.** Pictures are examined by an image model and voice messages are transcribed and screened as text, but both are newer and weaker than the text screen. They miss things, can fail without anybody knowing, and must not be relied on.
- **Call audio and video are not screened.** A call between two adults usually goes straight between their devices; every other call goes through Naven’s own call server, and only when that cannot carry it through Cloudflare’s relay or, last, Agora, a separate company with its own [privacy policy](https://www.agora.io/en/privacy-policy/). Nothing Naven runs screens, moderates or records them.
- **Age is self-declared.** Bands are set from a birth year given at sign-up. Naven does not currently run documentary age verification.
- **Messages are not end-to-end encrypted.** Pre-send screening and end-to-end encryption cannot both be true — content that never reaches a server cannot be screened. Naven is transport-encrypted throughout and chose screening, because it is built to be usable by children.

Messages, calls and communities in any browser. Free, with no advertising and no phone number.

- Hello[hello@navenapp.com](mailto:hello@navenapp.com)
- Help[help@navenapp.com](mailto:help@navenapp.com)
- Safety[safety@navenapp.com](mailto:safety@navenapp.com)
- Privacy[privacy@navenapp.com](mailto:privacy@navenapp.com)
- Schools[schools@navenapp.com](mailto:schools@navenapp.com)
- Partners[partners@navenapp.com](mailto:partners@navenapp.com)

© 2026 Naven · Nothing on this site is a promise; the [Terms](https://navenapp.com/terms) apply.

[Privacy](https://navenapp.com/privacy) · [Children’s privacy](https://navenapp.com/privacy/children) · [Terms](https://navenapp.com/terms)