How does meeting data flow?
The data flow for uploads and the MeinGPT notetaker — from capture through transcription to summary.
This overview distinguishes between uploading your own file and recording with the MeinGPT notetaker. In both cases, the transcript and summary are then processed in MeinGPT.
Note
EU capture and transcription: The hosted notetaker runs through MeetingBaaS in France, and speech-to-text processing runs through Gladia in the EU. The model used for a subsequent summary depends on the models approved for your workspace and their configured data region.
Data Flow Overview
Step 1: Capture source
You can upload an audio or video file or invite the MeinGPT notetaker to a supported online meeting. An upload is transferred directly to MeinGPT. The notetaker path additionally works as follows:
- MeetingBaaS (SAS SPOKE) provides the hosted bot that joins the meeting and records audio/video.
- It processes the meeting URL, bot metadata, and participant and speaker information.
- Speech-to-text transcription is provided separately through Gladia.
For both paths:
- 🔐 TLS-encrypted transfer to MeinGPT platform
- 🛡️ Organization-scoped storage in EU data centers
- 📋 Supported formats: MP4, MP3, WAV, M4A, MOV, AVI, WebM, AAC, FLAC, OGG, Opus
Note
Upload Limits:
- Audio files: Maximum 1 GB
- Video files: Maximum 5 GB
- Maximum duration: 4 hours 15 minutes
Video files are automatically converted to audio. The extracted audio must not exceed 1 GB.
Step 2: Ingest into MeinGPT
After a notetaker recording, MeetingBaaS reports completion through an authenticated callback, with polling as a technical fallback. MeinGPT copies the recording into organization-scoped storage. The audio track is used for transcription, while video can be made available for playback.
Step 3: Forwarding for Transcription
Transfer to Gladia (EU):
- 🇫🇷 Gladia - French company
- 🔐 Encrypted transfer of audio/video file
- 🎯 AI-based transcription in EU data centers
Step 4: Transcription & Return
Gladia processes your file:
- 🎧 Audio-to-text conversion with timestamp
- 👥 Automatic speaker recognition
- 📝 Structured transcript is sent back to MeinGPT
- 🧭 Processing under the agreed Gladia data and retention configuration
Step 5: Meeting storage
In the MeinGPT database:
- 💾 Encrypted storage of recording, transcript, and summary
- 🔐 Access according to your workspace permissions and any shares
- 📍 Storage location: EU data centers
Step 6: AI Summary
Summary is created:
- 🤖 An AI model approved for your workspace
- 📊 Extraction of key points and action items
- 💾 Storage of summary in MeinGPT DB
Step 7: Your Control
You have complete control:
- 👀 View transcript and summary
- 🗑️ Deletion of individual meetings possible at any time
- 📤 Export of the protocol as a Word document (DOCX); summary and transcript can be copied
🏢 Service Providers Used
| Step | Provider | Role | Region |
|---|---|---|---|
| Platform and customer storage | MeinGPT | Meeting management, permissions, playback, transcript, and summary | 🇪🇺 EU |
| Bot participation and recording | MeetingBaaS (SAS SPOKE) | Hosted MeinGPT notetaker for joining and recording meetings | 🇫🇷 France |
| Transcription | Gladia | Speech-to-text and speaker diarization | 🇪🇺 EU |
| AI summary | Approved model provider | Creates the selected meeting summary | According to model configuration and subprocessor list |
The current legal provider overview is available in the subprocessor list.
🔐 Data Protection & Security
Important security measures:
- 🔐 TLS-encrypted transfer between the involved services
- 🗝️ Authenticated MeetingBaaS callbacks with a per-bot secret
- 🧱 Organization-scoped storage and permission checks in MeinGPT
- 📋 Data processing agreements with the involved providers
Attention
Anyone inviting the notetaker or enabling automatic join must ensure the required legal basis, participant notice and, where necessary, consent before recording starts.
🗑️ Deletion & Your Rights
Your options:
- Delete individual meetings: Initiate deletion through the user interface
- End of contract: Return and deletion follow the agreed DPA, including its export and deletion process
Your GDPR rights: access, rectification, erasure and data portability apply to your meeting data just as they do to every other data category at MeinGPT — the full overview is in the security FAQ.
Note
The capture and transcription path is documented for the EU. For the subsequent AI summary, the data region of the approved model applies.
📞 Further Information
Data protection details:
- 🏛️ Data Protection Officer: datenschutz@heydata.eu
- 📋 Technical protection measures: Data Protection Documentation
Optional automatic follow-up emails
When an admin enables follow-up emails, MeinGPT creates a summary from the completed transcript and the stored speaker names – using GPT-6 Sol, hosted in the EU on Microsoft Azure. Private meeting notes are not included.
Recipients are the invitees of the calendar event, filtered by the organization's setting and the meeting owner's own setting. Each person receives an individual message through MeinGPT's email service. The email does not grant access to the recording.
The meeting owner can see recipients and delivery status in the meeting. Recipients can stop further follow-ups from the sending organization using the preference link in the email. See Automatic follow-up emails.