Cloud Horizon AI / Use cases
The LLM API for regulated EU teams.
Healthcare, financial services, public sector. EU-only inference, Dutch processor, no US parent in the chain. Built for compliance leads who get the second call after legal says no to OpenAI.
Six workflows your privacy office can sign off on
The bottlenecks specific to regulated work. Each one needs an LLM that does not leak inputs, does not phone home, and does not need a Schrems II carve-out.
PHI-safe clinical summaries
Pain
Your clinical team wants discharge summaries drafted from notes. The hospital privacy office will not let unstructured patient text leave the data centre, never mind cross the Atlantic.
Fix
Inference in Amsterdam under a Dutch DPA, no training on inputs, audit log per request. Pseudonymise on the way in, the model never sees direct identifiers, you put names back at render time.
KYC adverse-media triage
Pain
The compliance team reads hundreds of articles per onboarding case. Most are noise. The handful that matter get buried.
Fix
Stream every article through a classifier prompt with your risk taxonomy. Model returns a relevance score, a category, and a one-paragraph summary. Analysts see a pre-ranked queue.
Public-sector tender drafting
Pain
Drafting an answer to a 200-page TenderNed RFI is two weeks of senior bid manager time. Most of it is plumbing, the actual win-themes are five pages.
Fix
Long-context model ingests the RFI, your past wins, and your capability statements. Returns a section-by-section first draft you edit down. The bid manager spends time on win-themes, not on copying boilerplate.
Audit evidence drafting
Pain
External auditor asks for 80 pieces of evidence. Half are screenshots and policy excerpts you have written ten times before.
Fix
Model drafts the narrative paragraphs from your policies and recent change records. You attach the screenshots, sign off, ship to the auditor. Halves the prep cycle.
Regulatory change monitoring
Pain
EU AI Act, DORA, NIS2, MiCA, the new omnibus simplification package. Your compliance lead reads everything and forgets half of it.
Fix
Daily ingest of EUR-Lex, ESMA, EBA, AFM and BaFin feeds. Embeddings index against your control library. Diff drops into a digest with the affected controls flagged.
Patient or citizen self-service triage
Pain
Public services and patient portals get the same fifteen questions every day. Phone lines back up, email queues grow, the front-line team burns out.
Fix
Citizen-facing chat over your published policy and FAQ pages. Hard-coded refusals on anything that needs human judgement, escalation to a person on the rest. EU residency, Dutch DPA, GDPR DPIA template included.
Compliance posture, framework by framework
Where we sit today, what is on the roadmap, what your auditor will ask for. Architected for these frameworks. Certification claims are in our docs, not on this page.
| Framework | Region | Posture today | Notes |
|---|---|---|---|
| GDPR | EU/EEA | EU-only inference, Dutch processor, signed DPA | Article 28 processor agreement on team plans |
| Schrems II | EU | No US parent, no Cloud Act exposure | Spot Cloud B.V. is Dutch, no transfer to third countries |
| BSI C5 | DE | Frankfurt inference region, type 2 controls in scope | Evidence pack on enterprise plan |
| ISO 27001 | Global | Architected against Annex A controls | Statement of Applicability mappable per workload |
| EU AI Act | EU | Article 50 transparency notices, no prohibited use cases | Risk-tier classification helper in the docs |
| DORA | EU FS | Subcontractor register, exit plan, ICT risk reporting | Critical-third-party register entry available |
| NIS2 | EU | Incident reporting hooks, supply-chain register | Designed for essential and important entities |
| HDS | FR | Health data hosting roadmap, AWS Frankfurt today | HDS-certified region in 2027 H1 |
Per-request controls your auditor will love
Region pinning, audit tagging, PII redaction, retention overrides. Set them in the request header or body, every call carries its own policy.
# KYC adverse-media classification, PII redaction on by default
curl https://api.cloudhorizons.ai/v1/chat/completions \
-H "Authorization: Bearer $CLOUD_HORIZONS_KEY" \
-H "Content-Type: application/json" \
-H "Cloud-Horizons-Region: eu-ams-1" \
-H "Cloud-Horizons-Audit-Tag: kyc-case-784512" \
-d '{
"model": "glm-4.6",
"messages": [
{"role": "system", "content": "Classify the article against the risk taxonomy. Return JSON: {category, severity, summary, source_quality}."},
{"role": "user", "content": "<article body>"}
],
"redact_pii": true,
"log_retention": "30d",
"response_format": {"type": "json_object"}
}' Region pinning
Force a request to stay in Amsterdam or Frankfurt. 422 if the model is not available in that region.
Audit tag
Free-text label per request, surfaces in the audit log. Match to your case ID, your patient pseudonym, your bid number.
PII redaction
Inline redaction before the model sees the prompt. Names, emails, IBANs, BSN, NHS numbers, IP addresses.
Retention overrides
Default 30 days for prompt and response logs. Set log_retention: "0d" for zero retention on sensitive workloads.
Why regulated buyers pick us
A Dutch processor your DPO has heard of
Operated by Spot Cloud B.V. (KvK 89708873), a Dutch company under Dutch law. Standard EU Data Processing Agreement on team plans, no Cloud Act exposure, no parent in the United States. Inference in Amsterdam and Frankfurt, your data leaves the EU only if you explicitly opt in.
Join the waitlist