Skip to content
Atelyea product of Murat Gözel

AI Transparency Notice

Version
1
Effective date

You have probably arrived here from a message that told you a conversation is assisted by AI. This page explains what that means, in plain language and in about two minutes.

It is written for the person on the other end of the conversation — the customer messaging a business — not for a lawyer.

What the AI does

Businesses use Atelye to help answer the messages they receive on Instagram, Facebook Messenger and WhatsApp.

When you message one of those businesses, the software reads your message together with information the business has provided about itself — its products, its opening hours, its policies, its usual answers — and writes a suggested reply.

That is the whole of it. The AI drafts. It does not decide anything about you, it does not price anything for you, and it does not send anything.

A person checks every reply before it is sent

No reply reaches you until a person at the business has read it and approved it. The AI's suggestion goes into that business's inbox, where a member of staff reads it, edits it if it is wrong or does not sound like them, and then sends it.

If the software is not confident, or the subject is sensitive, it does not suggest anything at all and hands the conversation straight to a person.

Talking to a person

You can ask for a person at any time. Say so in the conversation — "I'd like to talk to a person", or anything that plainly means it — and the conversation is handed to a member of staff. That happens automatically, and it happens regardless of how confident the software was about its own answer.

You never have to justify the request, and asking does not end the conversation or lose its history.

Sensitive subjects, and anyone under 18

Some subjects never get an AI-drafted reply at all. They go straight to a person, every time, regardless of how confident the software is:

  • anything involving a child or a young person;
  • complaints, safety issues, and legal matters;
  • refunds outside the business's stated policy;
  • requests about your personal data;
  • an explicit request to speak to a person.

This is deterministic — it is not a judgement the AI makes about whether a subject is sensitive enough. Matching one of these triggers removes the AI from the reply.

What the AI sees

To draft a reply, the software uses:

  • the messages in your conversation with that business, including images, audio, video or files you have sent;
  • the identifier the messaging platform gives you for that business — an opaque code, not your real identity, and different for every business you message;
  • the business's own information — the description of itself, its products and its policies that it has written.

What it does not have: your profile, your activity elsewhere, a purchase history from other businesses, or anything about you from another conversation. Atelye does not build a profile of you, and it does not link the identity you have with one business to the identity you have with another. That is a deliberate design decision, not an omission.

Who processes your messages

Atelye is a platform operated by Murat Gözel, a sole proprietor established in Türkiye. The servers holding conversations are in the European Union.

To draft a reply, the content of the conversation is sent to a third-party AI provider — Anthropic, OpenAI or Mistral AI, depending on the business's configuration. Those providers process it to generate the draft. They are not permitted to use it to train their models.

Your messages also pass through the messaging platform you used — Instagram, Messenger or WhatsApp, all operated by Meta — because that is how they reach the business in the first place.

The complete list, with each party's role and location, is at /legal/subprocessors.

Conversation content is never used to train AI models, never sold, and never used for advertising. It is used to answer you, and to help the business you contacted answer people better over time — in that business's account only, never across businesses.

How to object, and how to have your data deleted

The business you messaged decides how your data is used, and Atelye acts on that business's instructions. So the first and quickest route is to ask that business directly — to stop using AI-drafted replies with you, to speak to a person, or to delete your messages.

You can also write to privacy@atelye.net. Atelye will pass the request to the business concerned and help them act on it, normally within 30 days. Atelye will not delete a business's records on its own initiative, because they are that business's records, not Atelye's — the one exception is where Instagram, Messenger or WhatsApp sends a deletion instruction after you remove the app's access, which Atelye acts on directly.

The Data Deletion page explains the routes and what happens to the data.

Marking of AI-generated text

The EU Artificial Intelligence Act, whose transparency rules apply from 2 August 2026, requires providers of systems that generate text to mark generated content in a machine-readable way where that is technically feasible.

Atelye's position, stated openly because it is a judgement rather than a settled matter:

  • Atelye's output is short conversational text sent through Meta's messaging APIs. Those transports carry the message text and discard anything attached to it, and there is no established, interoperable technique for marking a chat message that survives them.
  • The underlying models are third-party general-purpose AI systems whose own providers carry marking obligations for what they generate.
  • Every message is read and approved by a person before it is sent, so a human takes editorial responsibility for each one.

Atelye therefore relies on clear disclosure that you can read — this page, and the message at the start of the conversation — rather than on machine-readable marking of chat text. This is a reasoned position on a question that is not settled. It will be revisited as practice and guidance develop, and it would have to be revisited if replies were ever sent without a person approving them first.

Roles under the EU AI Act

For completeness, since these terms appear in the legislation:

  • Atelye is the provider of the AI system. Building the disclosure, the route to a human and the safeguards into the system is Atelye's obligation, and they are built in rather than left to configuration.
  • Each business using Atelye is a deployer. It decides how to use the system with its own customers and remains responsible for what it sends you.

A business can change the wording and the language of the disclosure message. A business cannot switch the disclosure off, and cannot remove your route to a person. Those are structural.

The disclosure message

This is the default text inserted at the start of an AI-assisted conversation. A business may reword it to sound like itself; it may not remove it.

English

This conversation is assisted by AI. A person from our team checks and approves every reply before it is sent. Ask to talk to a person at any time and we will pass you to someone. More: https://atelye.net/legal/ai-disclosure

Turkish

Bu görüşmede yapay zekâ desteği kullanılmaktadır. Gönderilen her yanıt, iletilmeden önce ekibimizden bir kişi tarafından kontrol edilip onaylanır. Dilediğiniz an bir kişiyle görüşmek istediğinizi belirtebilirsiniz. Ayrıntılar: https://atelye.net/tr/legal/ai-disclosure

Contact

Questions about this notice, or about how your data is handled: privacy@atelye.net

Murat Gözel, trading as Atelye — Akarca Mah. Fatma Seher Hanım Cad. Yuvam Akarca B1-B1 No: 98 İç Kapı No: 9 41310 İzmit Kocaeli Türkiye — +90 532 590 2141