> For the complete documentation index, see [llms.txt](https://help.steeple.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.steeple.com/en/assistant-ia/assistant-ia-confidentialite.md).

# Privacy, security, and permissions

This article describes what the AI assistant can see, what it can do, and how your data is processed.

## Respect for user rights

The assistant acts **always within the rights of the logged-in person**. It cannot:

* Access content to which the user does not have access in Steeple
* Carry out an action that the user could not perform themselves
* Bypass the roles, permissions, or moderation rules in force

## Confirmation of sensitive actions

Any action that **creates, modifies, or deletes** a piece of data triggers a **request for explicit confirmation** :

* A clear summary of the intended action is displayed
* Two buttons (**Yes** / **No**) allow you to accept or refuse
* No write action is carried out without validation

Simple viewing actions (search, reading) do not require confirmation.

## Data sent to the model

To respond to your message, the assistant sends the following to the AI model:

* The content of your message
* The files you attached (extracted text, image)
* The information strictly necessary for the request (for example, the name of a colleague being searched for)
* The context of the current conversation

No data is used to train models.

## Conversation storage

* The **current conversation** is saved so you can resume it after a reload or navigation
* Only one conversation is kept at a time per organization
* Starting a new conversation deletes the previous one
* Attached files are not **stored** beyond the processing of the message

## Improvements and quality

The feedback **thumbs up / thumbs down** you leave on the assistant's responses is used for evaluation and quality improvement purposes. It may be accompanied by an optional comment.