Providers

elmtree chat uses LLM inference providers for hosted model responses. These are technical service providers that execute a specific request against a language model and return the answer.

We work exclusively with LLM inference providers that meet two requirements:

1. They do not store chats as their own chat history.
2. They do not analyze chats for training, advertising, profiling, or comparable purposes.

What we send to providers

A provider receives the request needed to generate the model answer. We do not send an elmtree account ID, name, or email address. The request is not transmitted in a way that lets the provider associate it with your elmtree account.

If you use pseudonymization, the provider sees the pseudonymized version with tokens instead of the detected original values.

What providers should not receive

Providers should not receive a permanent chat history from us. They should not store, analyze, or use your chats to train their own models.

This is a central selection criterion for providers we use in elmtree.

What you should still consider

Even with strong provider rules, an LLM call is a transfer to an external service. Do not include personal data that is not needed for the task. For documents with sensitive information, use the preview and pseudonymization.

For especially confidential content, data minimization is stronger than trust in policies. Remove unnecessary passages, pseudonymize personal data, and disable web search when it is not needed.