What is the Terra Humana Assistant?
The Terra Humana Assistant is an AI-powered assistant of Fondazione Terra Humana ETS. This page transparently explains how it works, what data it processes and where its limits are — in accordance with Art. 50 of the EU AI Act.
What information is it based on?
The assistant answers exclusively on the basis of the Foundation's public documents: the Statute, RUNTS registration data, project descriptions and frequently asked questions.
Every sentence of an answer must be backed by a verbatim quote from one of these documents. The cited sources are shown below each answer. If the public documents do not contain a reliable answer, the assistant says so openly instead of making something up.
How does it work technically?
Your question is matched against the Foundation's public documents (semantic search). The most relevant passages are then passed to a language model (LLM), which formulates the answer while quoting the sources word for word.
Providers used: Microsoft Azure OpenAI Service (primary model), Anthropic (fallback model in case of outages) and Voyage AI (semantic analysis of the question).
What data is processed?
The text of your question is transmitted to the providers named above to generate the answer. Questions are not used to train models.
In the Foundation's internal logs we do not store question texts in plain text — only anonymised hash values, technical data (language, response times) and counters for the usage limit.
Legal basis: the Foundation's legitimate interest in informing its members efficiently (Art. 6(1)(f) GDPR). Data controller: Fondazione Terra Humana ETS, Località Sa Medada, 08044 Jerzu (NU), Italy. To exercise your rights (access, rectification, erasure, objection) write to info@terrahumana.ngo.
What are the limits?
Answers are generated automatically and may contain errors. Only the Foundation's official documents are authoritative.
The assistant does not provide legal, tax or medical advice and does not make decisions. For binding matters, always contact the Foundation's governing bodies.
What about the Telegram bot @TerraHumanaBot?
The Foundation's Telegram bot (@TerraHumanaBot) is an AI assistant as well: it uses models from Anthropic (Claude), OpenAI (GPT) and Google (Gemini) to answer messages, recognise receipts and transcribe voice messages. The same principles as on this page apply: answers are generated automatically, may contain errors and do not replace the Foundation's governing bodies.