Last updated: 26 August 2026
This policy explains what people are allowed to write on Bonded, what happens when something crosses the line, how to tell us about it, and what we will do. It is the longer version of the Community Rules you see in the app, and it applies to everyone who uses Bonded.
It should be read together with the Terms of Service, the Privacy Policy, and the Law Enforcement Request Guidelines.
1. There is very little to moderate, on purpose
Most apps ask you to trust their moderation. We would rather give you less to worry about in the first place. Bonded was built so that there are only a handful of places where anyone can write anything at all.
Here is the complete list of what a person can put on Bonded:
- Their username. Seen by everyone they connect with, and in search.
- Their avatar, built from a fixed set of pieces we design. Nobody can upload a picture.
- A short bio, up to 200 characters, seen only by people they have accepted as a connection.
- Answers to prompts we write, up to 300 characters each, again seen only by accepted connections. The prompt itself is written by us; the answer is theirs.
- A note attached to a report, which only our safety team ever reads.
Everything else you see in Bonded — the prompts, the vibes you can send with a ping, the colours and feelings you can attach to an answer — is wording we wrote and you choose from. There is no free text hiding in any of it.
There is no messaging on Bonded. You cannot send another person words of your own, anywhere in the app. A ping carries a feeling we wrote, not a sentence you typed. This is a deliberate design decision, not a feature we have not got to yet.
Nobody can upload anything. No photos, no images, no voice notes, no video, no files. If we ever change that, we will rewrite this policy before the change reaches you, not afterwards.
How the writing that does exist is handled
Bios and prompt answers are checked on our servers, before they are saved, against a list of slurs and abusive terms, and against anything that looks like a link. Usernames are checked the same way. If something is caught, it is never stored and never shown to anyone.
That check is deliberately literal. It catches the obvious, which is most of what it needs to catch. It does not attempt to judge tone, intent, or the meaning behind a sentence.
Separately from it, and not at the moment you press save, we can run a second check that does look at meaning. Bios, prompt answers and usernames may be read by a language model, which we ask to find writing that breaks the rules in section 2. We start it deliberately; it is not watching as you type, and it does not run on every save.
What matters about it is what it cannot do. It cannot remove anything, hide anything, or suspend anyone. All it can produce is an item in the queue described in section 3 — the same queue your reports go to, read by the same people, on the same deadlines. Every decision in section 5 is still made by a person who has read the thing themselves. A machine’s opinion has never removed a sentence from Bonded and, as this policy stands, it cannot.
We do this because reporting alone has a gap in it: a prompt answer is seen only by people who already accepted you as a connection, so something written to make one person uncomfortable may be seen by nobody else who would report it. Checking the writing itself is how that gap gets closed.
Which companies can process that text is listed in the sub-processor list; they are pinned to a named set, and configured so that none of them keeps what is sent or trains on it. Text that has not changed since it was last checked is never sent again.
A prompt answer is also different from a message in a way that matters: it is written for nobody in particular, it is never delivered to you, and it never appears in a notification. It sits on a profile until someone who is already a connection chooses to open it and read it.
Free accounts can hold up to four (4) answers; Premium accounts up to ten (10). The filtering, the audience, and the reporting route are identical either way. Paying for Bonded does not relax a single rule on this page.
2. What is not allowed
These are the Community Rules, in full:
- Being under 18. Bonded is for adults only.
- Harassment — repeated, targeted, or intimidating contact.
- Threats of harm to anyone, including to themselves.
- Hate or discrimination based on who someone is.
- Unwanted sexual content or advances.
- Impersonating another person.
- Sharing someone else’s private or confidential information.
- Spam, scams, phishing, or asking people for money.
Breaking these rules can cost you access to Bonded. Section 5 explains what we actually do, and in what order.
3. How to tell us
You can report:
- A ping — open the ping card, tap the menu, choose Report.
- A person — go to Friends, open the menu next to their name, choose Report.
- A specific profile answer — go to Friends, tap the person, choose “Report this answer”.
You can also email safety@bonded.studio at any time, about anything, including things that have nothing to do with a single ping or answer. Authorities should use the same address and read the Law Enforcement Request Guidelines.
Everything lands in the same place. Reports from the app, emails to safety@, and anything the automated check in section 1 flags all arrive in one queue, are worked by the same people, and are held to the deadlines in section 4. We do not keep a second queue, and an automated flag does not jump ahead of a person’s report — with the single exception in section 4 that applies to any report, however it arrives.
An item raised by the automated check is labelled as such where our team sees it, along with how confident the check was. That is deliberate: a moderator should know they are looking at a machine’s guess and not at somebody’s account of what happened to them, before they read it rather than after.
Reporting someone and blocking them are two different things. When you file a report, we ask whether you would also like to block that person. If you say yes, it takes effect immediately, before anyone at Bonded has looked at anything. If you say no, the report still reaches us. We ask rather than assume, because “someone should look at this” is not the same as “I never want to hear from this person again” — and flagging one bad sentence from a friend should not cost you the friendship.
The person you report is never told who reported them.
If you report the same thing twice, it does not count twice. But reporting someone’s answer and separately reporting the person are two genuinely different complaints, and both will be reviewed.
4. How quickly we respond
The clock starts when you file the report, not when we happen to open it.
| How serious | What it covers | We respond within |
|---|---|---|
| Urgent | Anything involving a minor; threats of harm | 24 hours |
| Severe | Harassment, unwanted sexual content, hate or discrimination | 24 hours |
| Everything else | Impersonation, privacy, spam, other | 72 hours |
Urgent reports do not wait to be spotted. They trigger an immediate alert to our safety team, which deliberately carries no user content in it — only enough to find the report in a system that is locked down and access-controlled.
If we are ever unable to review reports for a stretch of time — illness, or anything else — we switch off the parts of Bonded where people write text, rather than leave them running unwatched. Reporting stays available the whole time.
5. What we can do about it
In roughly increasing order of severity. Whatever we choose, we record what we did and why.
- No action — we looked, and no rule was broken. This is still written down, not thrown away.
- A warning to the account.
- A username reset, where the username itself is the problem.
- Removing a single answer, where one prompt answer is the problem. The answer comes down and the slot is freed; the rest of the profile and the account are untouched. This is the right answer to most reports about profile answers, and we reach for it first: one unacceptable sentence is not a reason to end someone’s account.
- A rate restriction — fewer pings.
- A ping suspension — the account keeps working, but cannot ping anyone.
- Account suspension — temporary, while we look further.
- A permanent ban — the account can no longer sign in and disappears from search for everyone.
Anything involving a minor skips all of this. A credible finding means an immediate permanent ban and, where it applies, a report to the relevant authority.
We tell you when we act on something you wrote. If we remove an answer, reset a username, or restrict or suspend an account, the person affected is told what we did, which rule it was under, and what they can do about it — that is section 7. Two things we include on purpose:
- Whether a person or the automated check in section 1 first flagged it. You are entitled to know a machine found it, even though a person decided what to do. We are not going to describe a human review as if it started with a human report.
- That a decision can be appealed, and how.
We do not tell you who reported you. That has not changed.
6. What we keep, and for how long
- The record of a report — who reported, who was reported, the reason, what we decided, and when — is kept as our permanent moderation record. It is what lets us see a pattern of behaviour over years rather than treating every complaint as the first one.
- A copy of the reported content itself is kept for 90 days after we have reviewed the report, and for 180 days if it has not yet been reviewed. It is held longer when unreviewed on purpose, so that a backlog can never quietly destroy the evidence before anyone has read it.
- Pings are visible in the app for 24 hours and are then deleted.
What people write on Bonded never appears in our analytics, our error reports, or our ordinary system logs. The Privacy Policy explains in full what we hold about you and the rights you have over it.
7. If you think we got it wrong
If we take action against your account, you can appeal by writing to safety@bonded.studio.
- Wherever we are able to, your appeal is reviewed by someone other than the person who made the original decision. Bonded is a small team, so this is not always possible. Where it is not, the decision is re-examined from the evidence against this policy, and we say plainly in our reply that this is what happened.
- We answer appeals within 14 days.
- The outcome, and the reasoning behind it, is recorded alongside the original report.
As a very small company, we are not legally required to run a formal complaints process. We offer one anyway, because banning someone with no way to argue back is not defensible whatever the law permits.
8. Law enforcement and emergencies
Requests from police and other public authorities are handled under our Law Enforcement Request Guidelines.
Where there is a credible and immediate risk to someone’s life, we may pass information to the relevant authority without waiting for a formal request. We record every time we do this, and why.
9. Transparency
Once a year we publish a report covering: how many reports we received and of what kind, what actions we took, how quickly we responded, how many appeals we received and how they turned out, and any orders we received from authorities.
It separates reports filed by people from items raised by the automated check in section 1, and for the automated ones it states how many were upheld once a person had looked. That last number is the one worth reading: a check that flags a great deal and is upheld rarely is a check making work rather than finding harm, and publishing it is how we hold ourselves to noticing that.
10. Changes to this policy
We review this policy at least once a year, and immediately whenever something meaningful changes — what people can write, what can be reported, how long we keep things, or how our team is staffed.
Some changes require us to rewrite this page before they reach you rather than after:
- Any kind of answer other than text — voice, video, images, or files of any kind.
- Profile answers becoming visible to anyone beyond an accepted connection.
- Any feature that lets one person send another person words of their own.
- Any automated system being allowed to remove, hide or restrict something on its own, without a person having read it first. The check described in section 1 cannot do this today, and turning it into something that can would be a change to what this page promises rather than an improvement to how it is enforced.
If any of those ever happen, you will be able to read the updated policy first.
Questions, reports, or appeals: safety@bonded.studio