Acceptable Use Policy

Prohibited content

Some things are never acceptable here, regardless of intent, artistic framing, or the fact that a model produced them rather than a camera. There is no research exemption and no satire exemption for anything on this list.

• Content that sexualises minors, in any form, style or claimed age
• Unlawful content, or content that facilitates serious harm
• Hateful content targeting people for who they are
• Content designed to harass, threaten or intimidate a specific person
• Sexual content depicting a real, identifiable person without their consent
• Content presenting fabricated events as genuine footage of real people
• Instructions or assistance for weapons, serious violence, or attacks on people or systems

Impersonation and likeness

Do not create personas or media depicting a real, identifiable person without their consent or another lawful basis. This includes public figures, and it especially includes cases where the result could reasonably be mistaken for real.

Satire and commentary have latitude; deception does not. The test we apply is not whether you meant it as a joke but whether a reasonable person, seeing it out of context, would believe it happened.

Voice counts. A cloned voice is a likeness, and consent for a face is not consent for a voice.

Deception and provenance

Do not use generated media to deceive people about the world: fake news events, fabricated evidence, invented endorsements, false testimonials, or a synthetic person presented as a real customer.

Do not strip or falsify provenance to hide how something was made, and do not claim a generated result is unedited footage. Where a platform or a law requires an AI disclosure, make it.

Safety and guardrails

Do not attempt to bypass identity, safety or spend guardrails. That includes prompt-based circumvention, and it includes routing a request through an agent to do something you could not do directly. An agent acting on your instruction is you acting.

Probing for weaknesses in order to report them responsibly is welcome and we would rather hear from you than not. Probing in order to exploit them is not, and the difference is entirely what you do next.

Other people’s rights

Do not upload or generate content that infringes someone else’s intellectual property, and do not upload material you do not have the rights to use as a reference. Owning a prompt is not owning the thing it describes.

Do not upload other people’s personal data: photographs of identifiable people who have not agreed, private documents, or anything you would need someone’s permission to hold.

Platform integrity

No scraping, no abuse of the interfaces, no automated account creation, and nothing that degrades the service for other creators. Load that would be rude if a person did it by hand is not acceptable because a script did it instead.

One person, one account. Do not create accounts to evade a suspension, to claim a benefit more than once, or to inflate a referral.

Financial conduct

Do not use stolen payment methods, do not launder value through credit purchases, and do not resell platform access or credits. Credits are tied to your account and are not transferable.

Referral abuse (self-referral, fabricated accounts, incentivised sign-ups that do not reflect real interest) is covered by the Referral Terms and reverses the commission it earned.

Enforcement

We may remove content, restrict features, or suspend and close accounts that break these rules. Severity and repetition both matter: a first mistake and a pattern get different answers, and we would rather correct than punish.

Some things do not get a warning. Content on the prohibited list, and conduct that puts a real person at risk, is acted on immediately.

Where we can explain a decision without exposing whoever reported it, we will.

If we get it wrong

If you think we got it wrong, say so through support and a person will look again. Bring the context. Most disputed decisions turn on something we could not see from the content alone.

A moderation decision does not touch your credits, your billing or the work in your library. Losing access to a feature is not losing your work.