# How we keep your keys and data safe

> DigitSpace stores your API keys only as hashes and your provider keys encrypted, keeps request metadata but not prompts unless you turn retention on, and runs in Singapore. You control retention, guardrails, IP ranges, budgets and sign-in for your organisation.

Source: https://digitspace.ai/security · Last checked: 2026-10-30

## How secrets are stored

| What | How it's stored | Who can read it |
| --- | --- | --- |
| Your DigitSpace API keys | A one-way HMAC-SHA256 hash; the key is shown once | Nobody, including us |
| Your provider keys | Encrypted with AES-256-GCM | Only the gateway, at the moment it calls the provider |
| Passwords | A scrypt hash | Nobody |
| Cards | Held by Stripe; we keep only a reference to the card | Stripe |

## What we keep about requests

| Data | Kept | Why |
| --- | --- | --- |
| Model, provider, tokens, cost, timing, status | Yes | Your usage history, billing and routing health |
| The prompt and the answer | Only if your organisation turns retention on | So you can review requests in your dashboard |
| Guardrail matches | Which check matched and what it did; never the text | So you can see what was blocked or redacted |

Retention is off by default.

## Controls you have

- **Data retention:** Off by default; your organisation decides whether prompts and answers are kept.
- **Guardrails:** Block or redact secrets, personal data, prompt injection and your own terms before a request leaves DigitSpace.
- **IP ranges:** Allow or deny calls by IP range on a key, a member or a team.
- **Budgets and access rules:** Cap spend and limit models and providers per key, member and team.
- **Sign-in:** Passwords, passkeys, and single sign-on for Enterprise organisations.
- **Audit log:** Changes to keys, members, billing and settings are recorded with who made them.

## Where it runs

DigitSpace is a Singapore company. The gateway, API and database run on AWS in Singapore. Requests are sent on to the AI provider you or routing choose, which processes them under its own terms; the sub-processors page lists every company involved.

To report a security problem, write to hello@digitspace.ai with the subject "Security report". We read every report.

## Questions

### Do you train models on my data?

No. DigitSpace doesn't train models. Providers handle requests under their own API terms; most major providers don't train on API traffic by default.

### Can your team read my prompts?

Not unless your organisation turned retention on, and even then our admin tools don't show prompt content.

### What happens if a key leaks?

Delete or roll it in the dashboard and it stops working at once. A budget and an IP range on each key limit what a leaked key could do.

### Do you have a SOC 2 report?

Not yet. Write to us if you need a security questionnaire answered for your review.

## Related

- [Guardrails](https://digitspace.ai/features/guardrails)
- [Budgets and access](https://digitspace.ai/features/budgets-and-access)
- [Enterprise](https://digitspace.ai/solutions/enterprise)
- [FAQ](https://digitspace.ai/faq)

Start free: https://digitspace.ai/signup · Docs: https://digitspace.ai/docs · Pricing: https://digitspace.ai/pricing