Trust Center
Subprocessors
Overview of all subprocessors engaged by DeepMask GmbH to provide the DeepMask platform. The focus is on German and European infrastructure: the primary hosting partner is the Schwarz Group's sovereign German cloud (STACKIT). All subprocessors have been carefully selected, are bound by contract under Art. 28 GDPR and do not use customer data to train AI models.
Service
Cloud infrastructure, compute & Kubernetes, object storage, managed databases and STACKIT AI Model Serving (LLM inference for Qwen3, GPT-OSS 120B, Gemma 3 27B)
Certifications
Purpose of processing
Primary hosting partner: operation of the DeepMask platform and execution of AI models in a sovereign German cloud. Data does not leave Germany.
Service
Web hosting and DNS for the public web presences (website, Trust Center)
Certifications
Purpose of processing
Provision of the corporate website and associated DNS services; no processing of chat or project content.
Service
Secure EU AI inference layer for selected models (including DeepSeek V3.1, MiniMax M2.5, GPT-OSS 120B)
Purpose of processing
Execution of open-weight models on EU infrastructure so that prompt and response data does not leave the EU.
Service
Azure OpenAI Service (GPT-4o, GPT-4.1, GPT-5.x, o3-mini)
Safeguards
EU standard contractual clauses, EU-US Data Privacy Framework
Certifications
Purpose of processing
Provision of AI inference for OpenAI models. Inputs are not used to train models; processing takes place in the EU data zone, i.e. exclusively in EU regions.
Service
Amazon Bedrock (Anthropic models: Opus 4.5/4.6, Sonnet 4.5/4.6, Haiku 4.5)
Safeguards
AWS DPA with EU standard contractual clauses, AWS European Sovereign Cloud commitments
Certifications
Purpose of processing
Provision of AI inference for Anthropic models. Under the DPA, processing takes place exclusively in EU regions; inputs are not used for model training.
Service
Google Vertex AI (Gemini 2.0 Flash, Gemini 2.5 Pro/Flash)
Safeguards
Google Cloud Data Processing Addendum, EU standard contractual clauses
Certifications
Purpose of processing
Provision of AI inference for Gemini models with contractually assured data storage in the selected region; inputs are not used for model training.
Service
Mistral models (Mistral Large 3, Mistral Medium 3)
Certifications
Purpose of processing
Provision of AI inference for Mistral models on European infrastructure.
Service
AI-assisted web search (optional feature, only when actively used)
Safeguards
Certifications
Purpose of processing
Real-time research with source citations. Only the search query is transmitted, and only when users switch on web search in the chat. The platform issues that query pseudonymously under its own identity, with no user or account reference.
Service
Google Meet and Google Workspace for customer communication and demo appointments
Safeguards
EU standard contractual clauses, EU-US Data Privacy Framework
Certifications
Purpose of processing
Conducting video meetings and business correspondence; no processing of the platform's chat or project content.