This page says what AI on Artur.Work does with your data, in the order that matters: what we hold, who we send it to, how long we keep it, and what you can make us do about it.
The controller is the one-person business that operates https://ai.artur.work. Write to [email protected] about anything on this page, including a request to see, correct or delete your data.
The legal grounds are: performing our contract with you (your account, your generations, your payments), our legitimate interest (preventing abuse, keeping our records, answering refund questions), a legal obligation (invoices and tax records), and your consent where you opt into something, such as connecting a chat bot.
When you run a model, the request is sent to the provider that serves it. The provider receives what the request contains - your prompt and any file you attached - and not your name, e-mail or account identity. Which provider will serve a request is shown before you send it.
The providers currently reachable from here are: Anthropic, BlackForestLabs, GoApi, Google, Ideogram, Luma, Mesh, NanoBananaApi, OpenAi, OpenRouter, Recraft, Runware, Runway, XAi.
What each provider’s terms say about training on what it receives, as we last read them. Where a provider lets us ask per request, we ask on every request unless you untick “Ask not to train” for it on your profile.
| Provider | Training | Terms |
|---|---|---|
| Anthropic | Does not train on API data | Reviewed 2026-09-27 |
| BlackForestLabs | Trains, no opt-out | Reviewed 2026-09-27 |
| GoApi | Not yet reviewed | |
| Does not train on API data | Reviewed 2026-09-27 | |
| Ideogram | Does not train on API data | Reviewed 2026-09-27 |
| Luma | Not yet reviewed | |
| Community computers (Mesh) | Community computers: cannot be enforced; use trusted rigs | |
| NanoBananaApi | Not yet reviewed | |
| OpenAi | Does not train on API data | Reviewed 2026-09-27 |
| OpenRouter | Trains unless asked not to | Reviewed 2026-09-27 |
| Recraft | Does not train on API data | Reviewed 2026-09-27 |
| Runware | Not yet reviewed | |
| Runway | Does not train on API data | Reviewed 2026-09-27 |
| XAi | Does not train on API data | Reviewed 2026-09-27 |
Each provider has its own terms and its own privacy policy, and some of them are outside the European Union. We rely on the data-protection terms in each provider agreement for those transfers. If that matters to you, restrict your account to providers that do not train on user input, in your profile settings, and check the provider before you send.
Decisions (/v1/decisions) are answered by TypeSafe's Jev model, which we reach through OpenRouter: both receive the text and questions of the request, nothing else. We ask OpenRouter to route only to endpoints that do not keep or train on what they receive (data_collection: deny). We do that unless you untick “Ask not to train” for OpenRouter on your profile. When you leave the model on Auto, the words of your request also go to Jev the same way, to tell what kind of work it is - never your files, only measurements taken from them, such as text found in an image. Those calls always ask not to keep or train, and are not made at all when your settings exclude OpenRouter.
If you send work to the peer network, it runs on another user's computer. That person's machine receives the prompt and produces the result. Turn the network off in your profile settings if you would rather not.
Generated and uploaded media is stored on our content network and served from an address derived from the contents of the file. Such an address cannot be guessed, but it is not protected by a login: anyone you give the link to can open the file, and so can anyone they pass it on to, until the file is deleted.
We do not sell your data, we do not share it for advertising, and we do not use it to train models.
Two cookies: one that keeps you signed in for the session, and one that remembers you if you asked to stay signed in. There is no advertising or analytics tracking on this site.
You choose how long your messages and images are kept, including keeping nothing at all. The full table, and what stays behind whatever you choose, is on What we keep, and for how long.
This page changes when the service does. Changes are published here rather than e-mailed, so that using the service does not mean agreeing to receive mail you did not ask for.
Last updated: 2026-09-23. The English version of this page is the one that governs. The other languages are translations offered for convenience; where they differ, the English text applies. Questions about this page: [email protected].