AI data residency for companies: what to require from an LLM provider
AI data residency has three parts: where your prompts are stored, where the model runs (inference), and whether retries and fallbacks stay in the same region. A provider can offer the first and not the other two. Standard Compute covers inference, retries and fallbacks with a Europe rule you set for the company, a team or a single key.
Create a company workspaceTalk to usSee pricing
This page lists the questions to put to any AI vendor, what OpenAI and Anthropic offer on their own APIs, and how the Europe rule works on Standard Compute.
AI data residency requirements, part by part
What to ask each vendor, and the answer on Standard Compute.
| Requirement | What to ask | Standard Compute |
|---|---|---|
| Inference location | What to ask: Where does the model run for our requests? | Standard Compute: With a Europe rule, every request is served only by approved European inference endpoints |
| Retries and fallbacks | What to ask: Can a retry or failover land in another region? | Standard Compute: No. Retries and fallbacks stay under the Europe rule, and failover never leaves your region or provider rules |
| No eligible endpoint | What to ask: What happens if no endpoint in the region can serve the model? | Standard Compute: The request is refused before your prompt is sent |
| Prompt storage | What to ask: Are prompts written to disk or logs? | Standard Compute: No. We process prompts in memory and never write them to storage or logs |
| Account and usage records | What to ask: Where is metadata kept? | Standard Compute: Frankfurt, Germany. Request records hold model, provider, tokens, cost and status, never prompts or outputs |
| Entry point | What to ask: Where does TLS terminate? | Standard Compute: Our EU gateway terminates TLS and runs in Finland |
| Scope | What to ask: Can the rule differ by team or app? | Standard Compute: Yes. Set it for the company, a team or a single key |
| Evidence | What to ask: Can we show which provider served each request? | Standard Compute: The request audit export lists the model, the provider that served it and the region rule in force, kept for 12 months |
Data residency on the OpenAI and Claude APIs
Both large model vendors now offer residency controls on their own APIs, with different coverage. The table summarises their own documentation, checked on 18 November 2026.
| OpenAI API | Claude API (Anthropic) | |
|---|---|---|
| Europe | OpenAI API: Europe (EEA and Switzerland) at eu.api.openai.com, with regional storage and regional inference | Claude API (Anthropic): Inference geo offers global (default) and us only |
| Storage at rest | OpenAI API: Regional storage for supported endpoints; system data is excluded | Claude API (Anthropic): Workspace geo us is the only option |
| How it is set | OpenAI API: Per project, chosen when the project is created by an eligible organization; the region can't be changed afterwards | Claude API (Anthropic): Per request (inference_geo) or as a workspace default and allow-list |
| Price | OpenAI API: 10% uplift for models released on or after 5 March 2026 | Claude API (Anthropic): US-only inference at 1.1 times the standard rate on Claude 4.6 and later |
On Amazon Bedrock and Google-managed Vertex AI, Anthropic says the region comes from the endpoint or inference profile. See Azure AI Foundry, Bedrock and Vertex AI data residency.
Residency routing across many models
Residency gets harder when a company uses more than one model vendor. Each vendor has its own setting, its own regions and its own exceptions, and a fallback to a second vendor can sit outside the rule you set on the first.
On Standard Compute the rule sits on the key. Set a Europe rule for your company, a team or a single key, and every request, including retries and fallbacks, is served only by approved European inference endpoints. If no European endpoint can serve it, the request is refused before your prompt is sent. Claude Opus 5.5, Sonnet 5.5 and Haiku 5.5, GPT-6.1 Sol and Terra, and open models such as GLM, DeepSeek and MiniMax all have European endpoints.
Add a provider allow-list ($199 a month, or in Enterprise) to choose exactly which inference providers may handle your data. Lower levels can only narrow the rules, and new providers are never added without an admin's approval.
Data residency and GDPR
GDPR does not require that personal data stay in the EU. It requires a lawful basis, a processor contract (Article 28) and safeguards for transfers outside the EEA (Chapter V). Residency is how many companies make the transfer question simple: if the processing happens in Europe, there is less to assess.
Sign our Data Processing Agreement in the dashboard in minutes. It commits us to no training on your data, no retention of your content, and breach notification within 72 hours. We publish every sub-processor with its purpose, data and location, and give you 14 days' notice before adding a new one.
Questions from IT, security and finance
- What is AI data residency?
- The commitment that your prompts and outputs are stored and processed in a region you choose. For an LLM it covers storage, the location where inference runs, and whether retries and fallbacks stay in that region.
- Does the Claude API offer inference in Europe?
- Anthropic's data residency page lists two inference geos for the Claude API,
globalandus, andusas the only workspace geo for stored data. On Standard Compute, Claude Opus 5.5, Sonnet 5.5 and Haiku 5.5 are served by approved European endpoints under a Europe rule. - Do retries and fallbacks respect the region?
- On Standard Compute, yes. Every request, including retries and fallbacks, is served only by approved European inference endpoints under a Europe rule, and refused before sending if none can serve it.
- Can different teams have different residency rules?
- Yes. Set the Europe rule for the whole company, for a team or for a single key. A lower level can narrow the provider rules above it, never widen them.
- What does the Europe rule cost?
- Region rules and zero data retention are free on every company workspace.
Related pages
- Azure AI Foundry, Bedrock and Vertex AI data residency
- European AI gateway
- GDPR compliant LLM API: what to check, API by API
- Private LLM for companies
- Sub-processors and locations
- Standard Compute for companies