AI API security & data residency (India)

India's AI teams face a real security puzzle: routing prompts through global APIs often means your account, billing, and support all sit outside India, which complicates DPDP compliance and incident response. unoblox keeps your account, keys, logs, and billing entirely on Indian infrastructure—encrypted, audited, and invoiced in rupees.

Data residency, honestly explained

For unoblox's self-hosted open models—Qwen, Llama, and Gemma—every step of inference runs on unoblox's own India-based GPUs (hosted with Yotta, a certified Indian data-center provider). For proprietary models you reach through unoblox—GPT, Claude, DeepSeek—your account, API keys, request logs, audit trail, and GST invoice all stay on Indian systems, while the token generation itself happens on that model's own provider infrastructure, exactly as it would if you called them directly. Either way, one Indian entity is accountable for your data handling, keys, and billing.

Security layers

Layer	Implementation	unoblox
Transport	TLS 1.3 with perfect forward secrecy	Always on
Data at rest	AES-256-GCM encryption	All logs encrypted
Auth	API keys, key-level isolation	ub-gw-… per key
Audit logs	90–180 day retention	Exportable, immutable
Billing & compliance	GST invoicing, DPDP-aligned handling	India-based throughout

Every invoice is GST-compliant and input-tax-credit eligible, so your security and compliance spend is also recoverable.

How to route through unoblox

1. Replace your provider's base URL with https://api.unoblox.ai/v1.
2. Swap in an unoblox key (ub-gw-…).
3. Confirm in the dashboard that logging and billing are running against Indian infrastructure.
4. Pull an audit export each quarter for your compliance filing.

No app rewrite needed. Claude, GPT, Qwen, Llama, and DeepSeek are all reachable through the same endpoint, same key.

DPDP-aligned data handling

unoblox follows data-minimization practices in the spirit of the DPDP Act: no unnecessary logging of personal data, configurable retention (90–180 days), and deletion on request via the API. This fits fintech, healthtech, and ed-tech workloads where user PII is core to the product—pair it with a model whose own data terms suit the sensitivity of your workload.

Frequently asked questions

Do you run my inference on shared servers?

For unoblox-hosted open models (Qwen, Llama, Gemma), yes—shared multi-tenant GPUs at Yotta India, isolated by API key and network-level controls. For GPT, Claude, or DeepSeek, your request is authenticated and logged in India, then computed on that provider's own infrastructure—the same trust boundary as calling them directly, minus the separate account and card.

Can I review your security posture before signing?

Yes. Yotta India can share its facility certifications directly, and unoblox provides its own DPDP-alignment documentation. Contact sales@unoblox.ai to set up a security review.

Do you store my prompts or responses?

Only for the duration of processing, plus your configured log-retention window (90–180 days), after which they're deleted. Nothing is used to train unoblox's own systems without explicit opt-in.

Can I self-host instead of using the API?

The unoblox gateway itself is cloud-only. If you need on-prem inference, you can self-host an open model like Qwen or Llama and still front it with an OpenAI-compatible layer.

How does rupee billing help with security audits?

Every invoice and payment receipt sits inside the Indian legal and tax framework, which keeps your audit trail simple—one GST-compliant record per month, no cross-border paperwork to reconcile.

What happens to data-handling terms if unoblox changes ownership?

Data-handling and residency commitments are contractual and carry over; there's no change to how your data is processed without advance notice.

Get started in rupees → https://unoblox.ai/sign-in