Integration / Model provider
Anthropic, routed by jurisdiction.
Reach Claude models through YTT Secunet on the deployment that matches each tenant's jurisdiction: Anthropic's own API for US-only inference, or Claude on Amazon Bedrock and Google Cloud Vertex AI in European regions.
Anthropic's first-party Claude API controls inference location with the inference_geo parameter. At the time of writing its documentation lists two values: global, the default, where inference may run in any available geography, and us, which keeps inference on US-based infrastructure. Workspaces can pin allowed_inference_geos and a default_inference_geo, and the response usage object reports where inference actually ran. Workspace geo, which governs storage at rest, is currently us only. There is no European inference geography on the first-party API, so a European tenant that must stay in the EU needs a different path to Claude.
That path runs through the hyperscalers. Claude models are offered on Amazon Bedrock, where geography-scoped cross-Region inference profiles (model ids prefixed eu.) keep requests inside AWS EU Regions, and on Google Cloud Vertex AI, which lists Claude in EU regions such as europe-west1 (Belgium) and europe-west4 (Netherlands) and an eu multi-region endpoint. Model availability differs by region and changes often, so YTT Secunet's endpoint registry checks the specific model and region pair rather than assuming parity.
YTT Secunet lets you keep one Anthropic-compatible integration while the policy decides which of these deployments a request may use. A US healthcare tenant can go to the Claude API with inference_geo set to us, injected by YTT Secunet so application code cannot forget it. A German tenant is routed to an EU Bedrock profile or a Vertex AI EU endpoint, using your own cloud credentials. Note one trap: Microsoft documents that when Claude models are used through Microsoft's generative AI services, customer data is processed by Anthropic in the United States, so YTT Secunet does not classify that path as EU in-region.
| Region or deployment | Location | Jurisdiction |
|---|---|---|
| Claude API, inference_geo us | United States | US / United States |
| Amazon Bedrock EU cross-Region inference profile (eu.) | AWS Regions in the EU | EU / European Union |
| Vertex AI europe-west1 | St. Ghislain, Belgium | EU / European Union |
| Vertex AI europe-west4 | Eemshaven, Netherlands | EU / European Union |
| Amazon Bedrock eu-west-2 | London, United Kingdom | UK / United Kingdom |
- 01
Decide per jurisdiction which Claude deployment is allowed: the Claude API with inference_geo us for US tenants, Bedrock or Vertex AI EU endpoints for European tenants.
- 02
For the Claude API, set allowed_inference_geos to ["us"] on the workspace used by YTT Secunet so the provider also enforces the restriction.
- 03
For Bedrock or Vertex AI, register your cloud credentials in YTT Secunet and confirm the exact model is enabled in the target region or EU inference profile.
- 04
Add each deployment under Sovereign Endpoints and attach the policy packs that may use it (for example us-hipaa-only, eu-gdpr-strict).
- 05
Point the Anthropic SDK at https://eu.yttsecunet.com (or the US data plane) and send the X-YTT-Policy and X-YTT-Tenant headers.
- 06
Check the receipt on a test call: the provider region id should read the Bedrock profile or Vertex region, never global.
import Anthropic from "@anthropic-ai/sdk";
const claude = new Anthropic({
baseURL: "https://eu.yttsecunet.com",
apiKey: process.env.YTT_SECUNET_API_KEY,
defaultHeaders: {
"X-YTT-Policy": "eu-gdpr-strict",
"X-YTT-Tenant": "tenant-fr-0087",
},
});
// Resolved by policy to an EU Bedrock profile or a Vertex AI EU region
const msg = await claude.messages.create({
model: "claude-sonnet",
max_tokens: 1024,
messages: [{ role: "user", content: "Classify this complaint under our escalation policy." }],
});- An EU policy never resolves to the first-party Claude API while its only inference geographies are global and us.
- On US routes YTT Secunet injects inference_geo us and rejects a response whose usage.inference_geo does not match.
- Claude served through Microsoft's generative AI services is classified as US processing, following Microsoft's own EU Data Boundary documentation.
- Bedrock EU profiles are static: AWS publishes a new profile id rather than adding Regions, so YTT Secunet pins the profile id in the receipt.
Does Anthropic offer an EU region on its own API?
Not at the time of writing. The documented inference geographies are global and us. EU residency for Claude is available through Amazon Bedrock and Google Cloud Vertex AI EU regions, which is what YTT Secunet routes European tenants to.
Do I need separate code for Bedrock and Vertex AI?
No. Your application speaks the Anthropic Messages format to YTT Secunet; YTT Secunet translates to the Bedrock or Vertex AI request format for the resolved endpoint and signs a receipt naming it.
Is a Bedrock EU profile the same as a single region?
No. A geography profile can process a request in any AWS Region in that profile, all inside the EU. If your contract or regulator requires one named country, pin a single in-Region deployment where the model is offered there, and YTT Secunet will refuse the profile.
Sovereign route ready
Stop risking 4% global turnover GDPR fines on AI inference.
Deploy sovereign, cryptographically proven AI routing in 15 minutes.