DATA RESIDENCY

Where your data lives

All customer data is stored in the EU by default. This page describes exactly which infrastructure services we use, where they run, and what options enterprise customers have for stricter residency requirements.

Standard infrastructure posture

By default, all customer data is stored and processed within the EU. The following services handle customer data.

Database	Neon PostgreSQL — Frankfurt, Germany (AWS eu-central-1). All application data, user data, audit logs, and AI outputs.
Application hosting	Vercel — compute region fra1 (Frankfurt). Edge cache nodes serve from nearest PoP; data is pinned to Frankfurt.
Transactional email	Resend — EU region (Dublin, Ireland). Handles login links, export notifications, and system alerts.
Error monitoring	Sentry — EU region (Germany). Stack traces and error context. PII scrubbing applied via beforeSend hook.
Product analytics	PostHog — EU cloud instance (eu.posthog.com, Frankfurt). User interaction events. No PII in event properties by default.
AI observability	Langfuse — EU cloud (Frankfurt). LLM call logs, prompt versions, evaluation results.

AI / LLM compute location

Anthropic has not yet released an EU-region API endpoint as of May 2026. We monitor release announcements and will update routing automatically when an EU endpoint becomes available.

Standard	Anthropic Claude (Haiku, Sonnet, Opus) via API — US region. Anthropic does not yet offer an EU-region API endpoint.
EU-only customers	Azure OpenAI Service — Sweden Central or Germany West Central. Available on request for customers with EU-only data residency requirements.
SCC coverage	Standard Contractual Clauses (SCCs) are in place for Anthropic (US transfer). Prompts contain only workspace metadata and application attributes — no user PII.
Model training opt-out	Enforced via API agreement with Anthropic. Customer prompt data is not used to train models.
Zero data retention	Configured with Anthropic API. Prompts and completions are not retained by Anthropic beyond the request lifecycle.
On-premises / BYOK	Planned — no runtime support yet. Enterprise customers may supply their own Anthropic or Azure OpenAI API key today (BYOL).

Data residency policy (Atlas product feature)

Atlas supports a per-workspace data residency policy with three tiers. This controls which AI providers can be invoked for that workspace.

eu_only

EU-only

All AI inference must use an EU-region endpoint. Anthropic (US) is blocked; Azure OpenAI EU is used instead. Enforced at the AI routing layer.

Available on request

us_ok

US OK (default)

AI inference may use Anthropic Claude (US region). SCCs and zero-retention agreements are in place. Suitable for most mid-market EA teams.

Default for all workspaces

on_prem

On-premises

All AI inference runs on customer-supplied infrastructure. Requires BYOK or private cloud deployment. Runtime support under planning — not yet generally available.

Under planning

Policy is configurable under Atlas → Settings → Workspace. All policy changes are logged in the workspace audit log.

Sub-processors

The full sub-processor list with data categories, regions, and DPA links is maintained separately.

View sub-processor list →

SCCs and GDPR-aligned transfers

Role	Spekir operates as a data processor under GDPR. You are the data controller.
SCCs	Standard Contractual Clauses are in place for all non-EU sub-processors (currently Anthropic, Inngest).
DPA	A Data Processing Agreement is available on request. Email hello@spekir.com.
Supervisory authority	Datatilsynet (Danish Data Protection Authority) — datatilsynet.dk
What we do not claim	We do not claim GDPR compliance as a certification — no such certification exists. We claim GDPR-aligned operations: documented processes, DPA chain, and right-to-erasure implemented.

Questions about data residency?

Procurement teams and legal reviewers often need a DPA, sub-processor notifications, or a custom residency configuration. We respond within two working days.

Talk to us about EA →