AI Policy

Where Gradiently uses AI, what is sent to the model provider, why nothing you make trains a model, and what you can rely on.

In effect from October 1, 2026Updated September 30, 2026
  • AI runs only when you ask it to.
  • Your prompts and designs are never used to train AI models.
  • What you make with the Designer is yours to use.
  • AI can be wrong. Check its work before you publish it.

8 sections

Two features use generative AI, and both say so where you use them:

  • The Designer makes and edits designs and Marks from what you describe, wherever you open it: Home, the Studio or the Forge.
  • Match a picture reads a picture you give it and can make new Marks in its spirit.

Everything else in Gradiently, including the Studio, the Forge controls, exports and the check that compares Marks, works without generative AI.

The models are provided by Anthropic. When you send a request, we pass on what the model needs to answer it:

  • What you typed, and the earlier messages in the same conversation.
  • The design or Mark you are working on, and your brand kit when it is relevant.
  • Any image you attach or ask it to match.

We don’t send your email address, password or payment details, and the model never sees other people’s private designs or uploads.

We don’t train AI models on your prompts, designs, uploads or Marks, and our agreement with our AI provider doesn’t allow it to either.

The provider may keep requests for a limited time to detect abuse, and then deletes them. The Privacy Policy covers this.

As far as we are concerned, what the Designer makes for you is yours, on the same terms as anything else you make in Gradiently. We claim no rights in it.

  • The Designer makes new work. It is built never to copy another person’s Mark, and a Mark too close to an existing one is refused like any other.
  • AI may give similar results to different people who ask for similar things. Only a claimed Mark is exclusive.
  • In some countries, work made mostly by AI may not be protected by copyright. We can’t promise that it is.

AI can misread a request, get facts wrong or place something badly. You decide what to keep: review text, names, numbers and claims before you publish, and make sure you have the rights to any picture or logo you give it.

The Acceptable Use Policy applies to AI requests too. Don’t use the Designer to imitate a real person or brand, to mislead people, or to make content that is unlawful or harmful. Requests that break the rules may be refused.

The Designer runs on credits. Every account starts with 30 free credits, and you can top up when you need more. A reply that fails because of a fault on our side gives its credits back automatically; a reply you stop yourself does not.

Credits you paid for and haven’t used can be refunded. See the Refund Policy.

You can connect an assistant such as Claude or ChatGPT to your Gradiently account over MCP or the API. It then reads and creates what your key allows, and spends none of your Gradiently credits.

That assistant is a separate service. What it does with the data it reads is governed by its provider’s terms and privacy policy, not ours. Give a key only the access it needs, and revoke it in Settings when you are done. The developer pages explain the scopes.

If the Designer made something it shouldn’t have, or you have a question about how AI handles your content, tell us or write to hello@gradiently.design. We read every message at gradiently.design.

Write to us. A person reads every request and replies by email.