Unknown inference location
Teams cannot always prove which GPU processed a request.
EU-contained AI routing
EU GPU or no inference.
EUBound is an OpenAI-compatible gateway that routes AI requests only to verified EU-located GPUs, blocks non-EU data paths and fallbacks, and returns signed proof for every inference.
Built for regulated European teams handling sensitive AI workloads.
The problem
Your AI traffic can leave approved boundaries without anyone noticing.
Prompts, completions, embeddings, files, logs, traces, caches, and fallback requests can move across providers and regions. For regulated teams, "probably hosted in Europe" is not enough.
Unsafe fallbacks
Reliability logic can silently route traffic outside approved regions.
Uncontrolled logs
Debug logs, traces, analytics, or support tooling can expose sensitive content.
No request-level proof
Compliance teams need evidence, not vendor marketing pages.
The solution
Route. Block. Prove.
EUBound turns AI data residency from a policy document into runtime enforcement.
Route
Send requests to approved EU AI providers through one OpenAI-compatible API.
Block
Deny any provider, model, fallback, or data path that violates your EU-boundary policy.
Prove
Return a signed sovereignty receipt showing where and how each inference was processed.
How it works
One API. Runtime policy. EU-contained route.
Request arrives
Your application sends a normal OpenAI-compatible request to EUBound.
Policy is checked
EUBound verifies allowed GPU countries, provider status, logging rules, subprocessors, and fallback boundaries.
Route is selected
The request is sent only to a verified EU-located GPU route that satisfies the policy.
Receipt is returned
The response includes signed proof of the route, policy, provider, GPU country, fallback status, and logging status.
Developer experience
Change the base URL. Keep your OpenAI client.
EUBound is designed as a drop-in OpenAI-compatible gateway for chat completions and embeddings.
from openai import OpenAI
client = OpenAI(
api_key="eub_...",
base_url="https://api.eubound.eu/v1"
)
response = client.chat.completions.create(
model="eu:auto",
messages=[
{"role": "user", "content": "Summarize this KYC file."}
],
extra_body={
"policy": "EU_HARD_BOUNDARY",
"receipt": True
}
)Request-level proof
Every inference returns a signed sovereignty receipt.
Each successful request includes evidence of the selected route, policy decision, GPU location, fallback status, logging status, and provider route.
{
"receipt_id": "rct_01JZ...",
"policy": "EU_HARD_BOUNDARY",
"decision": "allowed",
"provider": "scaleway",
"model": "eu:auto",
"gpu_country": "FR",
"customer_content_left_eu": false,
"fallback_used": false,
"prompt_body_logged": false,
"completion_body_logged": false,
"non_eu_content_subprocessor_used": false,
"evidence_hash": "sha256:...",
"signature": "ed25519:..."
}Runtime policy
Your AI policy should run in production, not live in a PDF.
EUBound evaluates every request against strict routing and data-boundary rules before inference.
EU_HARD_BOUNDARY
GPU must be inside an EU member state. Customer content must not leave the EU.
FR_ONLY
Inference is allowed only on verified GPUs physically located in France.
DE_ONLY
Inference is allowed only on verified GPUs physically located in Germany.
NO_BODY_LOGGING
Prompt and completion bodies are never stored by default.
Provider routing
Built to route across verified EU AI providers.
EUBound integrates with EU AI providers route by route. A provider is not automatically approved; each route must satisfy policy evidence before it can handle traffic.
- OVHcloud AI Endpoints
- Scaleway Generative APIs
- IONOS AI Model Hub
- Mistral routes with verified EU-contained processing
- Customer-hosted OpenAI-compatible endpoints
Use cases
For AI workflows that cannot leak across borders.
KYC and KYB
Summarize cases, extract missing information, and prepare analyst-ready files without uncontrolled AI data paths.
Legal and contracts
Route contract review and document analysis through verified EU-contained inference.
Insurance claims
Classify and summarize claim documents while enforcing EU-only processing.
Internal knowledge
Use AI over sensitive internal text while controlling provider routes and audit evidence.
Trust model
Built for legal, security, and engineering review.
EUBound does not claim to make your organization compliant by itself. It provides runtime controls and request-level evidence to support your internal GDPR, AI Act, vendor-risk, and security governance workflows.
No prompt-body logging by default
Audit metadata can be stored without storing customer content.
Fail-closed routing
If no approved EU route exists, the request is blocked.
Route-level evidence
Providers are evaluated per route, not blindly approved as a whole.
Signed receipts
Each inference can be tied to a policy version and evidence bundle.
Early access
Need AI inference that stays inside the EU?
We are opening design-partner conversations with fintech, regtech, legaltech, insurance, and European B2B SaaS teams handling sensitive AI workflows.