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

# Data

> How SynapsAI Cloud handles inference data, storage, and model artifacts

## Location

All models are hosted in T3/T4 data centers in North America. These facilities follow industry-standard physical and network security practices.

## Data processing

When you send data to your models:

1. All transmitted data and requests are encrypted using HTTPS.
2. Encrypted data is routed to a model instance on [shared GPU infrastructure](/isolation) for processing.
3. The response is returned through the same encrypted channel.

Data is never shared or sold to third-party services.

<Note>
  SynapsAI does not intentionally store or retain user data or requests beyond the duration necessary to complete processing. In-memory data associated with each request is cleared after processing, in accordance with our data handling policies.
</Note>

<Note>
  Generated videos remain in memory on our infrastructure (never written to disk) for up to 10 minutes after creation. Clients can delete generated videos immediately after download using the delete endpoint.
</Note>

## Model weights and artifacts

Model weights and prepared artifacts are stored on a secure internal filesystem accessible only to authorized servers within the same data center. These files are never exposed on the public internet.

## Your responsibilities

* Do not send regulated or sensitive data unless your compliance review permits it.
* Scope [API keys](/manage/api-keys) to minimum required models.
* Follow [Security best practices](/guides/security-best-practices) for key storage and access control.

For legal terms governing data processing, see the [Data policy](/legal/data-policy) and [Privacy policy](/legal/privacy-policy).

## Export and deletion

You can [export account data](/manage/account#export-your-data) or [delete your account](/manage/account#delete-account--danger-zone) from the dashboard.