Security
Security controls enforced in the request path.
Encryption, Postgres Row-Level Security, hashed virtual keys, in-path guardrails, and atomic credit reservation on every call.
How the request path works.
Every call is authenticated, scoped, and checked before reaching a provider. See the compliance page and trust center.
Virtual keys for every call.
Customer inference traffic uses scoped virtual keys. Provider credentials stay sealed on the server.
- SHA-256 hashed at creation; plaintext is shown once and never stored
- Per-key spend ceilings, RPM/TPM rate limits, and model allowlists
- Immediate revocation from the dashboard or API
Tenant isolation at the database.
Every tenant table enforces Postgres Row-Level Security scoped to your organization.
- Owner, Organization Admin, Member, and Viewer roles enforced in policy
- Cross-tenant queries return zero rows at the database layer
- Append-only audit trail records actor, IP, and timestamp on changes
Guardrails in the request path.
Pre-call and post-call safety checks run on every request, at no extra cost and behind no paid tier.
- Output PII redaction by default; input PII scanning per org, team, or key
- Prompt injection defense, secret scanning, and keyword filtering
- Blocked pre-call requests refuse cleanly and spend $0
Reserve and settle credit safety.
Every request reserves credits before dispatch and settles at the served model’s exact list price.
- Atomic credit mutations prevent negative balances under concurrency
- Held credits release automatically on upstream 4xx, 5xx, or timeout
- Stripe webhook ingestion verifies signatures and event idempotency
Encryption and observability.
Transport security, zero payload retention, audit logging, and the uptime target.
Encryption in transit and at rest.
TLS 1.2+ with HSTS preload on every endpoint and AES-256 encryption across Postgres, Redis, and object storage.
Zero payload retention.
Prompts and completions are processed in memory and never written to disk. Only billing and routing metadata is logged.
Audit-grade observability.
Append-only audit logs for administrative actions, key changes, and budget events with actor, IP, and timestamp.
Contractual uptime target.
Managed Azure Container Apps and managed Postgres deployment backed by the published Service Level Agreement.
99.9%
Pin where your data lives.
Default US footprint in Azure East US (Virginia), with regional pinning on Enterprise.
Default US footprint.
US East (Virginia) is the default region, on Azure East US, with encrypted managed Postgres replicated in one region.
Regional pinning on request.
Available on request for Enterprise: European Union (Netherlands), EU North (Stockholm), United Kingdom (London), Canada (Montréal), Australia (Sydney), Singapore, India (Mumbai).
No unrequested region moves.
Tenant data stays in its region unless you request a migration; EU subprocessors sign SCCs. Pin a region: sales@nrouter.ai.
Session and token policy.
Predictable timeouts, token rotation, and device revocation for dashboard sessions.
Idle timeout (30m default).
Inactive dashboard sessions warn at 25 minutes and require re-authentication at 30 minutes.
Refresh-token rotation.
Session refresh tokens rotate on every use, and replaced tokens are invalidated across active sessions.
Device management.
Account settings list active sessions by device, operating system, and IP address with one-click revocation.
Session vs API key separation.
Dashboard sessions authorize control-plane actions only. Inference calls require a scoped virtual key.
Report a vulnerability.
Coordinated vulnerability disclosure and critical incident escalation.
Email security disclosures
Send vulnerability reports to our security inbox. Acknowledged in < 48 hours.security@nrouter.ai · < 48 hoursCritical security phone line
Critical-only line. Voicemail is forwarded to on-call.+1 (650) 512-3902Coordinated disclosure
We follow coordinated disclosure and credit security researchers in our public changelog when permitted.
FAQs
Where is data stored, and can I pin a region?
All customer data lives in managed Postgres with at-rest encryption, replicated within a single region. United States is the default footprint. EU residency is available on Enterprise; the UK, Canada, Australia, Singapore, and India are available on request on Enterprise for residency-sensitive workloads. We do not move tenant data across regions without an explicit migration request.
Does nRouter store or retain LLM prompts and completions?
No. nRouter operates zero data retention for inference payloads. Prompts and completions are processed in memory for inspection and routing, then streamed directly to and from the provider. Payloads are never persisted to disk or database. Only operational telemetry (token counts, latency, list-price spend, timestamps) is retained for analytics and billing.
How does nRouter support HIPAA compliance and PHI?
nRouter supports HIPAA-eligible workloads on Enterprise. We execute Business Associate Agreements (BAAs), and when an input redaction rule is enabled, PII and PHI are redacted in-path before payloads reach upstream model providers.
Is nRouter SOC 2 certified?
Not yet — and we will not say "certified" until the report is signed. nRouter operates SOC 2-aligned controls today: encryption, access control, audit logging, change management, and tenant isolation, enforced from day one. A formal SOC 2 Type II observation period is currently underway. The underlying cloud infrastructure (Microsoft Azure) and managed database services are independently SOC 2 Type II certified. Customers who need assurance before the report lands can request a controls walkthrough or a completed vendor security questionnaire from security@nrouter.ai.
How does nRouter prevent cross-tenant data access?
Every tenant table has Postgres Row-Level Security enabled with policies scoped by organization, so a query against another tenant returns zero rows at the database layer.
What if my virtual key is leaked?
Virtual keys are tenant-scoped, rate-limited, budget-capped, and observable. Revocation is immediate from the dashboard or API; the in-memory cache invalidation propagates within seconds. Keys are hashed in the database. The plaintext is shown exactly once at creation. The blast radius of a leak is the budget on that single key, not your account.
Are AI guardrails included on every plan?
Yes. Built-in PII redaction, prompt injection detection, secret scanning, keyword filtering, and content moderation run on pay-as-you-go at 0% platform fee, with no paid tier gate.
How do I report a vulnerability?
Email security@nrouter.ai with details. We acknowledge reports in < 48 hours under coordinated disclosure and credit reporters in our changelog when permitted.