# Workspace region and data residency

> Where TalkDemo stores and processes your sessions, recordings and voice data, how the India region works, and what runs outside India.

Source: https://talkdemo.com/help/team-and-settings/workspace-region-and-data-residency · Updated 2026-11-20

## Key takeaways

- Each workspace has a data region. India (Mumbai) is the default; United States (US-East) can be chosen.
- The region decides where new sessions run and where their files are stored. A change applies to new sessions only.
- Indian-language voice is processed by Sarvam in India.
- AI model calls run in the United States, with personal data kept to what the task needs and no training on your data.

Your workspace's region decides where sessions run and where their data is stored. For Indian customers, that is India: calls, files and the database are in Mumbai, and voice is processed by Sarvam in India.

**Who can do this:** Admin changes the region. Everyone can see it.

## What the region controls

| Data or service | India region |
|---|---|
| Real-time call (audio and shared screen) | Routed to the region nearest the prospect: Mumbai for prospects in India |
| Cloud browser for live product mode | The provider's nearest region: Singapore for Indian sessions, as it has no India region yet. It renders your demo tenant, not prospect data |
| Recordings, screenshots, decks and files | AWS, Mumbai (ap-south-1) |
| Database: agents, sessions, transcripts, leads | Mumbai |
| Speech for Indian languages, Hinglish and English (India) | Sarvam, India |

## What runs outside India

- **AI model.** The agent's conversation, summaries and lead scoring use Anthropic's Claude in the United States. It receives the transcript, your knowledge passages and screen summaries, never demo credentials. Zero-retention settings are used where the provider offers them.
- **World-language voices.** Calls in English (US, UK, Australia), Spanish, French, German and the other world languages use Deepgram for listening and ElevenLabs for speaking, both in the United States. They are used only on turns in those languages, never for Indian languages. To keep all voice processing in India, turn on only Indian languages (see [Languages and voices](/help/agents/languages-and-voices)).
- **Embeddings.** Text from your knowledge sources, with no prospect data. Not sent at all while AI is turned off for your workspace, which also stops session summaries and narration drafts (lead scores then come from rules only); knowledge search then works by words inside TalkDemo.

The full list is on talkdemo.com/subprocessors.

## Check your region

1. Open **Settings**. **Workspace profile** opens.
2. Under **Data and tax**, **Data region** shows where your data lives: **India (Mumbai)** or **United States (US-East)**.

## Changing the region

An Admin can choose another **Data region** and click **Save data settings**. The change applies to new sessions only: existing recordings and data don't move, and the page warns you before you save. Talk to us before you switch, and before your first live session if most of your prospects are outside India. Moving existing data between regions is handled by our team, not in the app.

**Privacy:** keeping data in India supports your obligations under the DPDP Act. Your company remains the data fiduciary for prospect data; TalkDemo processes it on your instructions under the data processing agreement.

## Frequently asked questions

### Is a prospect outside India routed somewhere else?
The live call runs in the region nearest the prospect to keep delay low. What is stored afterwards, such as the recording and transcript, goes to your workspace's region.

### Do you offer other regions for Enterprise?
Data residency options are part of the Enterprise plan. Talk to us about your needs.
