How ArtSide's automated safety checkpoints work, what happens when content is flagged, what it means for your credits and account, and how to appeal a decision. The detailed rules live in our Acceptable Use Policy.

1.Why we run safety checks

ArtSide can generate and transform images from simple text descriptions and personal photographs. That creative power can be misused — to sexualise real people, to create deepfakes, or to produce other harmful or illegal content. To keep the service safe and lawful for everyone, every prompt, photograph and generated image passes through automated safety checks before it is processed, stored or shown.

These checks are automated. They are not a judgement about you: a large majority of blocked requests are false positives caused by an ordinary word or image tripping a filter. This page explains exactly what happens when something is flagged, and what you can do about it.

2.The three checkpoints

There are three automated checkpoints, matching the three ways content enters or leaves the service:

Your prompt — checked before a generation starts and before any credit is reserved for it (Text to Image, Photo Transformation, Photo to Art and AI Brush prompts are all checked).
Your photograph — any photo you upload is checked before it is registered and stored in your account.
The generated image — every image the AI produces is checked before it is saved to your account or appears in your gallery.

The checks are run by an automated safety classifier through our AI gateway. The classification requests are processed under a zero-data-retention configuration: they are handled in memory, not stored by the provider, and never used to train AI models. Our Privacy Notice describes the data handling in full.

3.What gets blocked

The checks screen against our Acceptable Use Policy. In summary, they look for:

any sexual content involving minors, real or synthetic — zero tolerance, with no appeal of the content decision itself;
non-consensual intimate imagery, including attempts to undress or sexualise a real person;
sexual deepfakes or sexualised depictions of real, identifiable people;
content promoting violence, violent extremism or serious unlawful harm;
unlawfully hateful content, harassment and targeted abuse;
attempts to deceive — fabricated evidence, forged documents, impersonation for fraud;
attempts to circumvent the safety system itself.

Ordinary creative, artistic, satirical and personal uses are welcome. Artistic nudity, dark themes, political satire and photographs of ordinary people are not blocked for simply being edgy — the checks target the specific harms listed in the Acceptable Use Policy.

4.What happens when something is blocked

If a prompt or upload is blocked, you will see a short message telling you the request could not be completed, without technical detail about why. If a generated image is blocked, it is never saved to your account and never appears in your gallery.

Your credits are safe: if a generation, edit or print preparation is blocked, any credits reserved for it are returned to your balance automatically. Credit handling is completely separate from safety decisions — you are never charged for blocked content.
Nothing is published: blocked content is not stored, shown or shared with other users.
Serious illegal content is deleted: where a check identifies suspected child sexual abuse material or similarly serious illegal content, the image is deleted immediately and permanently — we do not retain it — and the matter is referred to the appropriate authorities where the law requires or clearly justifies it.

5.Strikes, warnings and account restrictions

An automated block on its own does not put a strike on your account. A blocked prompt can be a completely harmless false positive, so every automated block is recorded as a moderation incident and reviewed by our team.

Strikes and other enforcement measures are applied only where a human reviewer confirms a violation, or where abuse is clearly repeated. Depending on the seriousness, outcomes range from no action (false positive), to a warning, a recorded strike, restricted features, temporary suspension, or permanent termination of the account for the most serious categories. Child safety and non-consensual intimate imagery result in immediate termination without warning.

You can always see the standing of your own account in your account area, and material enforcement decisions about your account involve human review.

6.What we record when something is flagged

We keep moderation records deliberately minimal. An incident record stores the type of check, high-level category flags (for example that a check was flagged as sexual or violent content), a severity tier, the time, and a short masked excerpt. We do not store your full prompt text or detailed classifier output in these records.

Incident records are deleted automatically after 180 days, unless they have been escalated for legal or security review. The Privacy Notice has the full picture, including your rights over this data.

7."The safety service is unavailable"

If a safety check cannot be completed — for example because the safety service is temporarily down or times out — the request is refused rather than allowed through unchecked. This is deliberate: we would rather briefly refuse a request than let content through with no check at all.

This is not a block against you and nothing is recorded as a violation. Simply try again in a few minutes, and any reserved credits are returned automatically.

8.How to appeal a flag

If you believe a prompt, upload or image was blocked in error, email [INSERT ABUSE REPORTING EMAIL] with: the approximate date and time of the attempt, which workflow you were using (Text to Image, Photo Transformation, Photo to Art or AI Brush), and a short description of what you were trying to create. Do not resend the blocked content in your first email — we can look it up from the incident record.

What happens next

A human reviewer looks at the incident record and the decision, never an automated re-check on its own.
If it was a false positive, we confirm the decision as dismissed, no strike is recorded, and where relevant we may adjust the underlying configuration so the same content is not blocked again.
If a violation is confirmed, we explain the outcome and any resulting strike or restriction, except in the most serious categories where we may not enter into correspondence.
We aim to respond to appeals within 5 working days, and account restrictions can be lifted where a review shows the flag was wrong.

Appeals and reports about other users' content both go to [INSERT ABUSE REPORTING EMAIL]. See the Acceptable Use Policy for what to include when reporting another user.

9.Known limitations

Automated moderation is imperfect. Some legitimate prompts and images will be blocked by mistake, and we cannot promise that every prohibited image is detected or prevented. The appeal route above is how we correct the first kind of mistake; the Acceptable Use Policy and our enforcement process are how we address the second.

Because ArtSide is in beta, the moderation configuration is still being tuned. If you hit a false positive, telling us about it genuinely helps improve the service.

Moderation FAQ

Contents