Principle
In client projects QLTY media uses only paid access whose contractual terms exclude training on submitted data. Which providers, regions and retention periods apply to your project is listed in the sub-processor register of the data processing agreement. Training with your data happens only with your written consent and only for your own model.
Sub-processors
| Provider | Service | Contract tier | Region | Retention |
|---|---|---|---|---|
| Meta | WhatsApp Business API | Business | EU/USA (standard data processing) | per Meta policy |
| OpenAI | API (business tier) | paid API, no training | USA | typically up to 30 days (abuse monitoring) |
| Anthropic | API (commercial) | paid API, no training | USA | typically up to 30 days |
| Mistral AI | Local operation (Ollama) | model weights local, no transfer | Germany (Cologne server) | none |
| Strato | VPS hosting | data processing agreement | Germany | project duration |
The project-specific list (which of these providers your project actually uses) is fixed in the DPA and updated on changes.
Technical and organizational measures (excerpt)
- Servers in Germany, in physically secured rooms
- Access limited to named persons; remote maintenance logged
- No transfer of project data to third parties
- Freelancers in the network are contractually bound to the same rules; consumer accounts for project data are prohibited
- Client data is not used to train external models (business tiers without training, local models without data transfer)
Contractual basis
Every AI project gets a data processing agreement under Art. 28 GDPR with listed sub-processors. We have run an exemplar contract in production since June 2026 (TVR 1898 e.V.). Chatbots in client projects disclose their AI nature (EU AI Act, Art. 50). The legal assessment of your use case rests with your data protection officer or legal counsel; we implement the technical requirements.
As of November 2026






