---
title: "Subprocessors — Mili"
description: "The subprocessors Mili uses to provide the Services, what each one handles, and how long it retains Customer Data. None may use Customer Data to train AI models."
url: https://getmili.ai/subprocessors
---

# Subprocessors

Mili uses the subprocessors below to provide the Services. Each is bound by a written agreement requiring appropriate safeguards, permitting use of Customer Data only to provide services to Mili, and prohibiting use of Customer Data to train or improve artificial intelligence or machine learning models. We can require deletion of Customer Data held by any subprocessor at any time.

## Model and transcription providers

These providers process Customer Data under zero-data-retention terms. Customer Data is used only to return the requested result and is not retained afterwards.

- **OpenAI** — language models. Queries, notes, transcripts and documents. No retention.
- **Anthropic** — language models. Queries, notes and transcripts. No retention.
- **Google Cloud** — language models. Queries, notes and transcripts. No retention.
- **AssemblyAI** — speech to text. Audio and transcripts. No retention.
- **Deepgram** — speech to text. Audio and transcripts. No retention.

## Processing services

These providers receive Customer Data to perform a specific task and do not retain it.

- **Thesys** — interface generation. Queries. No retention.
- **Parallel** — web research. Queries. No retention.
- **Perplexity** — web research. Queries. No retention.
- **ConvertAPI** — document conversion. Documents. No retention.
- **E2B** — sandboxed code execution. Queries. No retention.

## Infrastructure and storage

These providers hold Customer Data for as long as the customer firm’s own retention settings require.

- **Amazon Web Services** — hosting, storage and processing. All Customer Data.
- **Recall.ai** — meeting capture. Audio and transcripts.
- **Tiptap** — collaborative note storage. Notes.

## Operational tooling

These providers hold limited Customer Data for a defined period to support the reliability and quality of the Services.

- **Portkey** — model gateway and request logging. Transcripts and notes. Up to 30 days.
- **Braintrust** — model evaluation and tracing. Transcripts, notes and queries. Up to 30 days.

## Changes to this list

We update this page before adding a new subprocessor that handles Customer Data. To be notified of changes, write to [privacy@getmili.ai](mailto:privacy@getmili.ai). How Customer Data is handled is set out in our [privacy policy](/privacy-policy).