AI Acceptance Policy

The Service is a machine. It turns a text description into a new image, every time, using machine learning. No photograph of any person is used as an input, and no real person is depicted. The output is probabilistic: the same prompt will not produce the same image twice, and we do not promise that any particular prompt will produce a usable result.

We accept prompts describing adult human figures in a sexual context. We do not accept prompts describing minors, or prompts presenting the output as a photograph of a specific real person, in either case.

Before a prompt reaches the model, we run it through a screen. The screen refuses prompts containing:

• A description of a subject as child, under 18, or by reference to childhood or school age;
• The name of a specifically identifiable real person without a documented consent record;
• Words that describe non-consent, coercion, or force, presented as though real;
• A prompt longer than 2,000 characters, or a prompt that changes shape after we have seen it.

This list is the same as the one inthe Terms of Service, because it comes from the model licences we operate under. If we refuse a prompt, we will tell you in the UI; the screen’s decision is not itself an interpretation of your intent.

We run open-weight models whose licences we can verify. We do not send your prompt or image to a third party to generate it, and we do not use your content to train our own models unless we have said so and you have turned it on.

The output is a new image. You keep the rights you would normally have in it under your jurisdiction’s copyright law. The model’s own licence (CreativeML Open RAIL++-M or the equivalent in use on a given card) attaches to your use as well as ours.

In most jurisdictions an image produced wholly by a machine without human authorship is not, strictly, a “work” for copyright purposes. We make no representation that any image you generate will be protectable, or that it will not be similar to something else someone else has generated.

On request, we will tell you which model was used to generate a given image, when it was generated, and what prompt produced it. We will not keep the prompt’s words in our logs in a way that identifies the author beyond your account.