> ## Documentation Index
> Fetch the complete documentation index at: https://getlemma.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# HIPAA compliance

> How Lemma Intelligence protects patient health information: who can see it, how it's analyzed, how AI models handle it, and how it's audited.

Lemma Intelligence can answer questions about your remittances and claims, which contain protected health information (PHI). It handles PHI under the same [Business Associate Agreement](https://getlemma.com/business-associate-agreement) (BAA) your entity signs with Lemma during onboarding, and applies the safeguards below on top of the rest of the Lemma platform.

## It sees only the health data you can see

Lemma Intelligence reads your data as you, so it sees health data only if your role lets you see it in the dashboard, which takes the [Billing Team](https://getlemma.com/docs/guides/user-roles.md#practice-permissions) practice permission. An Owner can remove a member's access to health data at any time by turning off that permission. See [Control what it can see](https://getlemma.com/docs/lemma-intelligence/data-access.md).

## Patient identity stays out of analysis

When Lemma Intelligence calculates across your remittances, such as "denials by payer this quarter", the data it analyzes has no patient names, dates of birth, member IDs, or diagnosis codes. Each patient is an anonymous key, which is enough to count patients or group one patient's claims. The copy does hold what a biller looks a claim up by: claim numbers, dates of service, payer, rendering provider, procedure codes, and amounts.

That analysis runs on a temporary copy in an isolated sandbox with no network access. The copy holds only the tables the question needs, and is deleted as soon as the results are read. See [What you can do](https://getlemma.com/docs/lemma-intelligence/how-it-works.md#analyze-your-data).

## Patient names only on request

Patient names appear only when a billing team member asks about a specific remittance or specific claims, the same detail they'd see on the remittance page. Lemma Intelligence reads that remittance or those claims directly, with the member's own permissions.

## Audit trail

Every lookup that can return patient information is logged with who asked, when, which entity, and exactly which remittances and claims were returned. This record shows whose health data was read, for whom, and when.

## AI model vendors

Lemma Intelligence sends prompts to AI models through an AI gateway that Lemma has a Business Associate Agreement with. Every request is sent with zero data retention, so model vendors don't store your prompts or the data in them.

## What Lemma stores

* **Your conversations.** Questions and answers are stored encrypted, with the page each question was asked on. An answer can quote what it found, such as a payer's name or a claim number.
* **Query results are not stored.** Lemma keeps the queries a conversation ran, but not the rows they returned.
* **Logs never contain query text or results.** Lemma's logs record which tables a query read and how long it took, not what it asked or returned.
* **Rated answers and handoffs.** When you rate an answer, or hand a conversation to our support team, the conversation is shared with Lemma so we can review it or help you.

## Your responsibilities

* **Give health data access to the people who need it.** The Billing Team permission controls PHI access in Lemma Intelligence and across the dashboard, so grant it only to members whose job requires it.
* **Leave out entities you don't need.** Uncheck an entity in **Chat settings** to keep its data out of a conversation entirely.
* **Treat answers like any other PHI.** If you copy an answer that names a patient into another tool, that tool's handling rules apply.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.