AI Data Protection Information
    • PDF

    AI Data Protection Information

    • PDF

    Article summary

    Enterprise AI Security & Governance Disclosure


    1. System Overview

    FoxBase operates an AI-powered chatbot designed as a document-based informational assistant.

    Intended Use

    • Customer support

    • Sales support

    • Internal knowledge assistance

    The system:

    • Processes text input

    • Generates text responses

    • Retrieves information from provided documentation (PDF-based knowledge base)

    • Does not execute transactions

    • Does not access customer backend systems

    • Does not perform automated decision-making

    • Does not conduct profiling

    User-uploaded attachments are not supported.


    2. AI Model Architecture

    Model Type

    • Large Language Models (LLM)

    • Natural Language Processing (NLP)

    • Machine Learning / Deep Learning

    Supported Models (Customer Configurable)

    • Haiku (default)

    • GPT-4.1 / GPT-4.1 Mini

    • Sonnet

    • Gemini

    • Nova Micro / Lite / Pro

    FoxBase does not develop or train proprietary foundation models.

    Model Capabilities

    • Text-to-text generation

    • Contextual document retrieval

    • Semantic understanding of technical documentation

    The system does not support:

    • Image generation

    • Audio processing

    • Multimodal outputs

    • Autonomous actions


    3. Model Training & Data Usage

    • No model training is performed by FoxBase.

    • No fine-tuning of foundation models with customer data.

    • No prompt-tuning using stored conversations.

    • No ingestion of customer data into model training pipelines.

    Conversation data is not reused to alter model weights.


    4. Data Processing & PII Handling

    Personal Data Requirement

    The chatbot does not require personal data for operation.

    Possible PII Processing

    Personal data may be processed only if voluntarily entered by a user in text input.

    The system:

    • Does not actively collect personal data

    • Does not profile individuals

    • Does not evaluate individuals

    • Does not perform automated decisions with legal or significant effect

    Logging & Observability

    FoxBase uses observability tooling (e.g., Langfuse) for:

    • System monitoring

    • Debugging

    • Performance analysis

    • Retrieval optimization

    Stored elements may include:

    • User text inputs

    • Model outputs

    • Technical metadata

    Data is not used for model training.

    No automated masking or redaction layer is currently implemented.


    5. Infrastructure & Data Centers

    Core Hosting

    • AWS infrastructure

    • Region: Frankfurt (EU)

    Model Invocation

    Model routing depends on selected model configuration.

    Core infrastructure is EU-hosted.
    Model processing location may vary depending on provider configuration.

    FoxBase does not operate proprietary model infrastructure.


    6. Subprocessors

    Potential subprocessors include:

    • AWS (hosting infrastructure)

    • Langfuse (observability and logging)

    • Model providers (depending on selected model configuration)

    Data processing occurs under contractual agreements.
    Model providers process input data solely for response generation.

    No data is sold, shared for marketing, or used for independent analytics purposes.


    7. Security Controls

    FoxBase operates under an ISO/IEC 27001 certified Information Security Management System (ISMS), renewed annually.

    Security controls include:

    • Role-based access control (RBAC) via internal user management system

    • Access limitation to observability data

    • System logging and traceability

    • Organisational incident management procedures

    Incident Response Contact:
    support@foxbase.de

    Data Protection Contact (GDPR/DSAR):
    support@foxbase.de

    Encryption details (at rest / in transit) follow AWS standard security configurations.
    Formal encryption specifications can be provided upon request.


    8. Content Safety & Filtering

    FoxBase does not implement additional custom content filtering layers.

    Content moderation and safety controls are provided at the model-provider level, which typically include safeguards against:

    • Sexual content

    • Hate speech

    • Violence

    • Self-harm

    • Copyright violations

    FoxBase does not disable provider-level safeguards.


    9. Compliance & Regulatory Positioning

    GDPR

    • Data minimisation approach

    • No required personal data

    • No profiling

    • No model training with user data

    • Observability limited to system improvement

    EU AI Act

    The chatbot qualifies as a general-purpose AI system used for informational assistance.

    Based on current scope:

    • Not classified as a high-risk AI system

    • Subject to transparency obligations

    • No prohibited AI practices involved

    ISO Standards

    • ISO/IEC 27001 certified


    10. Change Management

    If system scope changes (e.g., transactional integration, automated decision-making), FoxBase conducts a reassessment of:

    • Data protection impact

    • AI risk classification

    • Regulatory obligations


    11. More Information

    Further security or data protection-related information is available upon request at support@foxbase.de.


    Was this article helpful?