---
title: "Data flow in MeinGPT chats"
description: "A transparent overview of how your chat data is processed in the MeinGPT platform"
canonical_url: "https://meingpt.com/en/docs/privacy-security/faq-chats-datenfluss"
language: en
---

# Data flow in MeinGPT chats

## How is your chat data processed?

A transparent overview of how your chat data is processed in the MeinGPT platform.

**Storage in the EU**: chat history and uploads sit in EU data centres. Where a
*prompt is processed* depends on the model you choose — see step 3.

## The data flow at a glance

### Step 1: your input

- 🔐 TLS-encrypted transfer to the MeinGPT platform
- 🛡️ Temporary storage in EU data centres
- 📋 Supported formats: text, PDF, image upload (jpeg, png, webp, gif), Word, PowerPoint, Excel, CSV, text file (txt), URLs
- 📦 Other formats (ZIP and other archives, audio, video, .msg/.eml) can be uploaded - the upload succeeds without an error. Whether the assistant can read the contents depends on whether the code sandbox is enabled in your workspace: with the sandbox, the file is placed in the assistant's workspace and can be unpacked and processed with code; without it, the contents are never read. Skill packages are separate - you upload those in the assistant editor under "Skills", where ZIP archives are always unpacked automatically.

### Step 2: PII filter (if enabled for your workspace)

- 🔍 Personal data is detected on meinGPT's own servers in Germany (Hetzner, Falkenstein)
- 🔁 It is replaced with placeholders before every request to a language model, for EU and non-EU models alike, and put back in the answer
- 📖 Overview: [PII filter](/en/docs/privacy-security/pii-filter); settings and limits in detail: [privacy filter](/en/docs/privacy-security/privacy-filter)

### Step 3: forwarding to the AI model

- 🇪🇺 Processing takes place at the provider of the model you choose. With DPA 2.0, all models run in the EU.
- ℹ️ The model provider and the place of processing are visible at any time in the model overview
- 🔐 Encrypted transfer of your message

### Step 4: processing & response

- 🧠 Text-to-text processing in the language model
- 🔄 Zero data retention: the model provider keeps no permanent copy of your input on its own infrastructure
- 📝 A structured response is sent back to MeinGPT
- ⚙️ Further processing by MeinGPT where necessary

### Step 5: storing the chat history

- 💾 Encrypted storage of the chat history
- 🔐 Access to your data is restricted by a role-based access concept
- 📍 Storage location: EU data centres (Germany)

### Step 6: your control

- 👀 Access to every chat history
- 🗑️ Individual chats can be deleted at any time

## Providers used

| Service   | Provider             | Location            | Data storage             |
| --------- | -------------------- | ------------------- | ------------------------ |
| Platform  | MeinGPT              | 🇪🇺 EU (Germany)     | Until deletion on your company's instruction; after the contract ends, return or deletion within 30 days |
| AI models | depends on selection | 🇪🇺 EU (location per model in the model overview) | No permanent copy at the provider (zero data retention) |

## Data protection & security

**At a glance:**
- 🚫 No permanent storage of prompts at the model providers
- 🔐 TLS-encrypted transfers, encrypted storage of the history
- 🗑️ After the contract ends, all of your company's data is returned or, on request, deleted within 30 days, including backups
- 📋 GDPR-compliant data processing agreements with every provider
- 🔒 Administrators release or block models organisation-wide

## Deletion & your rights

**Your options**:

- 🗑️ Delete individual chats: immediately, through the interface
- 🧹 Complete deletion: within 30 days after the contract ends if your company requests deletion rather than the return of its data, including backups; your company can also instruct the deletion of data at any time

**Your rights under the GDPR**: access, rectification, erasure and data portability apply to your chat data just as they do to every other data category at MeinGPT — the full overview is in the [security FAQ](/en/docs/privacy-security/faq-data-security).

**EU processing**: when you choose EU models, the model processes your chat
data in the EU. With DPA 2.0, all models run in the EU. Tools such as web
search call their own services. The [PII filter](/en/docs/privacy-security/pii-filter)
replaces personal data before any model sees it, wherever that model runs.

## Further information

- 🏛️ **Data protection officer**: datenschutz@heydata.eu
- 📋 **Technical safeguards**: [data protection documentation](/trust-center)
- 🔍 **AI providers & data protection**: [model overview](/en/docs/platform/models)