Privacy Notice

← back to Quite Human

Submitting an entry is for people aged 18 or over (see the terms). If you are under 18, please don't sign in.

What we store

Your Firebase Authentication account (Google sign-in — no password). We do not copy your email address into our own database.

While your submission is in moderation, we store the text you submitted and a link to your account. Once published, the text is stored with no link back to your account. Where a moderator takes a decision about a submission, we keep an internal record of it (see below) so decisions can be reviewed and enforcement can be applied consistently.

We keep a nightly backup, held in London for 30 days, of the published entries and of each account's record: its strikes and whether it has been banned. A backup deliberately leaves out everything that links an entry to the person who wrote it, including submissions still waiting in moderation, so keeping backups does not bring back a link that publication destroys.

We count submissions and reports per day to stop the site being flooded. To do that without keeping your IP address, the address is hashed together with a random salt generated fresh each day, and only the hash is stored, as a counter. The salt is never readable by anyone visiting the site, and it is destroyed along with that day's counters, so yesterday's counters cannot be joined to today's, or back to an IP address, once the day has turned.

Separately from anything we store ourselves, our hosting provider records ordinary web-server request logs, which include IP addresses, and deletes them automatically after a limited retention period.

What we never store

No password. No engagement data on published entries — there are no likes, replies, or view counts to collect in the first place.

No date of birth. Before you sign in we ask for it to confirm you are 18 or over, but it is checked in your browser and is never sent to us or stored, and neither is the fact that you passed.

Rejected and flagged submissions

Where a submission is not approved, we keep a record: the text you submitted, the decision taken, the reason the moderator gave, and the identifier of your account. The same applies when a published entry is later pulled after being flagged.

Every rejection is reviewed by a person before the submission is deleted, including a rejection the automated stages reached on their own. Until that review happens the submission stays in the moderation queue, still linked to your account. There is no deadline on it: a queue is worked through when there is somebody to work through it, so a rejected submission may sit there for some time. It is not on the site in the meantime, and it will not be unless the review overturns the rejection.

We keep this because enforcement depends on it. Strikes do not expire, and two strikes are a ban, so a decision taken today may still matter much later; without the record we could neither apply that consistently nor check whether an earlier decision was wrong. The record is also what lets us correct the automated classifier when it makes a mistake.

These records are readable only by the operator. They are never published, never shown to other users, and never shared with anyone else except where we are legally required to disclose them. They are not used for advertising, profiling, or any purpose other than moderation of this site.

Unlike published entries — where the link between the entry and the account that submitted it is deliberately destroyed at the moment of publication — moderation records remain linked to your account for as long as the account exists.

Decisions made automatically

Whether your submission is published is decided automatically. Three automated stages run in order: a list of prohibited terms, then OpenAI's moderation endpoint, then a classifier running on Google's Vertex AI. An entry that passes all three is published at the next nightly build, without a person having read it.

Rejection works the other way round. An entry the automated stages reject is kept off the site at once, but it is not discarded on a machine's word: a person reviews every rejection, and can overturn it. A person also reads a submission where the classifier is unsure, where one of the stages errors or returns something we cannot read, and where somebody flags an entry after publication. Nothing is ever published because a stage failed — a failure always sends the entry to a person, never onto the site.

No strike and no ban is ever applied automatically. The automated stages decide only whether a piece of text is published; anything that attaches to your account is decided by a person.

If a decision went against you and you think it was wrong, you can ask for a person to look at it again, and we will not refuse on the grounds that a machine made it.

Your rights

Under the UK GDPR you have the right to ask us for a copy of the personal data we hold about you, to ask us to correct it if it is inaccurate, to ask us to delete it, to object to or ask us to restrict how we use it, and to receive it in a portable form. Where a decision about you has been made by automated means alone, you can ask for a person to review it. You also have the right to complain to the Information Commissioner's Office.

To make any of these requests, email quitehuman.org@gmail.com from the address associated with your account. We will respond within one month. We may need to confirm that the request really comes from you before acting on it.

Two limits are worth stating plainly, because they follow from how the site is built rather than from any choice to resist a request:

Published entries usually cannot be traced back to you. The link between an entry and its author is destroyed when the entry is published. That is a privacy feature, but it means that if you ask us to delete "my published entry", we generally have no way to identify which one it is unless you can tell us. If you can, we will remove it.

Erasure of moderation records is not automatic. Where a record exists because of a strike or a ban, we may retain it after an erasure request, relying on our legitimate interest in preventing the ban from being trivially evaded by deleting the account and signing up again. We will tell you if we rely on this, and you can challenge it with us or with the ICO. Records not connected to any enforcement action will be deleted on request.

Contact

Email quitehuman.org@gmail.com for anything to do with this notice. The terms of service cover reporting an entry and appealing a decision about your account.

Who is responsible for your data

The operator of Quite Human is the data controller for the personal data described in this notice. The operator can be reached at quitehuman.org@gmail.com, which is the correct address for any question about this notice, any request under the rights listed above, and any complaint.

Where your data is processed

The site runs on Google Cloud Platform and Firebase. Account sign-in, storage of submissions, and the servers that run the site are all provided by Google.

The database and the servers are in the United Kingdom. Your account reference, your submissions while they are in moderation, and the moderation records described above are held in a Firestore database in London. The servers that receive your submission, check it and publish it are in London too, as is the file of published entries that the front page loads. Storing and handling a submission on our own systems does not take it out of the UK; the moderation services described below are the exception.

A few things do, or cannot be promised to stay. Firebase Authentication holds the account you sign in with, including the email address on it, and does not let us choose where; we cannot promise that it stays in the UK. The site's pages and images are delivered by Google's global network, so a visitor's request may be answered from outside the UK. And the ordinary web-server logs described above, which include IP addresses, are held in Google's logging service, which we have not restricted to the UK.

As part of moderation, and before any human sees it, submitted text is sent to two automated classification services. OpenAI's moderation endpoint processes it in the United States. Google's Vertex AI is called on its global endpoint rather than a region we pin, which means Google may process the text in any region where the model is available. Both act as processors on our instructions, and neither uses the text to train its models. Only the text of your submission is sent — never your account identifier, your email address, or anything else that identifies you.

Where processing happens outside the United Kingdom, we rely on the international transfer safeguards in our agreement with the provider concerned: for Google, its Cloud Data Processing Addendum, which incorporates the UK Addendum to the EU Standard Contractual Clauses, and equivalent terms for OpenAI.