Privacy and Processing
elmtree chat is built privacy-first: we want to process only the data needed for the selected function, and we deliberately separate chat, document processing, and RAG preparation.
For individual functions, you decide whether processing should happen locally in the browser or optionally through our servers.
Principles
We do not analyze chat data.
We do not store chats as readable server history.
We do not know your name.
No analysis of your chats
We do not analyze chat messages. In particular, we do not use your chat messages for LLM training, advertising, or profiling.
No storage of your chats
We do not store chats as readable server history: neither user messages nor LLM responses. The chat state remains local in your browser.
Anonymity toward LLM providers
We send requests to LLM providers so they cannot be associated with your elmtree account. We do not send your account ID, name, or email address to the provider.
Still, you should be careful with personal data in chat messages or documents. For sensitive topics, avoid names, email addresses, phone numbers, addresses, case numbers, or similar identifying data, or pseudonymize them before sending.
Documents only when you choose
Documents are only transferred to our servers when you use server-side processing for the task. Some functions, such as work with large documents, require server support. You decide whether to use them.
No analysis of your documents
We do not analyze uploaded documents. Not for training, not for advertising, not for profiling, and not for product analytics.
RAG only temporarily
RAG preparation can be stored temporarily in the server's local database to support work with large documents. This data is deleted again.
Main areas
- Document Processing explains when documents stay local and when our server can help.
- Pseudonymization explains how documents and related follow-up questions are redacted before sending.
- RAG Temporary Storage explains why prepared document data can be stored briefly.
- Data Minimization explains what we do not analyze and which metadata is still technically necessary.
- Providers describes our requirements for LLM inference providers.