SECURITY & PRIVACY

Where your content goes, and who can see it.

One page for your IT and legal team: which AI answers the questions, what the extension stores, and the controls you keep in TYPO3.

Three paths for inference.

You choose where queries and embeddings travel. All three can be used in the same TYPO3 installation for different sites.

  1. Local Ollama

    On your own server

    Visitor → TYPO3 → Local Ollama

    Stays on your server

    Questions, content snippets, and embeddings stay on infrastructure you run. Nothing is sent to an AI company.

  2. Managed Ollama API

    Hosted by Net2typo

    Visitor → TYPO3 → Net2typo EU Servers

    Hosted in the EU

    Open-weight Qwen3 models on EU infrastructure we operate. Your content is never sent to OpenAI, Google, or other AI companies.

  3. Your own AI key

    OpenAI, Gemini, Claude, OpenRouter

    Visitor → TYPO3 → Your AI provider

    Handled by your AI provider

    The question and the matching snippets go to that provider under your own contract and its data terms.

Six controls for enterprise data.

How the TYPO3 extension and backend isolate content, scrub secrets, and keep you in compliance.

  • Permission-aware indexing

    Frontend user-group restrictions are checked before a page or file enters the public index, so member-only content stays out of answers.

  • Isolated per site

    Embeddings, queues, and answers are tagged by site_root, so one installation keeps each site’s knowledge separate.

  • Consent before analytics

    Visitors are asked when a chat starts. Conversations are only logged when they agree.

  • Built-in input guard Starter plan and up

    Prompt injection and spam are blocked before a message reaches TYPO3 or the AI model.

  • Your privacy links and AI notice

    Link your own privacy policy and show an AI disclaimer in the chat widget, configured in the TYPO3 backend.

If we host the AI for you.

Safeguards of the Managed Ollama API, for sites that can’t run Ollama themselves.