Ethical AI.
Complete Transparency.
We build AI agents with ethics at the core. Here's exactly how we protect your data and ensure responsible AI use.
Our 3 Core Commitments
Simple principles that guide everything we do
Full Transparency
You always know when AI is used, where your data is stored, and how it's processed. No hidden practices.
You Control Your Data
Your personal data & knowledge stay on Magif.ai servers (EU). You can delete them anytime. Data sent to AI providers (OpenAI) is clearly disclosed before each agent creation.
EU AI Act & GDPR
We integrate EU AI Act and GDPR rules when appropriate, and educate our clients about their responsibilities. All data stored in EU (France & Germany).
Where Your Data Lives
100% stored in the European Union
Paris, France
Your Personal Information
Name, email, account details
NEVER used for AI training
Secured by AWS infrastructure
Berlin, Germany
Your Expert Knowledge
Books, articles, methods, content
Stored in Qdrant (open-source)
Never shared between clients
File Analysis (Google Gemini)
When you upload files (PDFs, documents), they are analyzed by Google Gemini to extract and structure your knowledge into a searchable format.
Agent Conversations (OpenAI)
When your AI agents interact with users, small relevant chunks of your knowledge base are sent to OpenAI using RAG (Retrieval Augmented Generation) to generate intelligent responses.
Magif.ai Anonymiser Engine
Military-grade privacy layer that protects your users' personal information in real-time
Smart Detection
Instantly identifies names, locations, and personal IDs with 99.3% accuracy
Encrypted Replacement
Replaces sensitive data with secure tokens, then restores it automatically
Instant Speed
Processes everything in under 50ms — completely invisible to users
How it works: Detects personal information → replaces with encrypted tokens → sends masked data to OpenAI → restores original context. OpenAI never sees your users' real personal information.
Questions?
We're committed to transparency and open dialogue
Report concerns, ask questions, or request our full policy documentation
Last updated: November 12, 2025 • Reviewed annually