startvisor.AI Logo

Digital Sovereignty & EU Procurement

Sovereign AI that clears the procurement gate — not just the demo.

AIonicOS is a single-tenant AI platform operated exclusively in the EU — aligned with BSI C5, the BSI C3A criteria and the EU Data Act. Built for public-sector buyers, regulated industries and sovereignty-conscious Mittelstand whose procurement now scores sovereignty as a weighted gate.

The 2026 procurement reality: sovereignty is a scored criterion, not a wish

Sovereignty is becoming measurable. The BSI C3A criteria (Criteria enabling Cloud Computing Autonomy), published in April 2026, let buyers assess cloud sovereignty objectively across six domains — from strategic and legal to technological autonomy. C3A is a voluntary BSI assessment framework, not a certificate, and it builds on the established, audited BSI C5 baseline and the EU Cloud Sovereignty Framework. Procurement teams are increasingly translating exactly these domains into weighted award criteria.

The rulebook is already live. The EU Data Act has applied since 12 September 2025 — including the Chapter VI switching and portability rights: a maximum two-month notice period, an obligation to remove switching obstacles, and a ban on switching charges from January 2027. Under the EU AI Act, GPAI, governance and transparency obligations have applied since 2 August 2025. Vendor lock-in is therefore no longer a commercial footnote but a regulatory risk that surfaces during qualification.

The core risk now has a name. The BSI lists 'Cyber Dominance' as a distinct threat category: the ability of market-dominant providers to retain permanent access to customers' systems and data. Add the third-country access exposure: the US CLOUD Act follows control over the provider, not the storage location — an EU data centre alone does not cure it. The EDPB and EDPS hold that providers subject to EU law cannot lawfully satisfy such disclosure requests. Sovereignty is decided by architecture, not by the map.

Six sovereignty dimensions — and how AIonicOS meets each one technically

Each row mirrors a domain that procurement officers and CISOs assess. On the left the procurement requirement, on the right the concrete AIonicOS mechanism — verifiable, not asserted.

EU data residency & inference
All data and every model inference must physically reside and run in the EU/Germany — at rest, in transit and during inference (GDPR Art. 44–49, BSI C5, Schrems II).
Single-tenant architecture: the full stack is deployed per customer on a dedicated EU host, with its own isolated Postgres/Apache-AGE, Qdrant, Valkey and SeaweedFS — no shared data plane. The residency gate (`residency.py`) enforces a strict provider allow-list under `data_residency=eu`, including a fully-EU path (Mistral FR + self-hosted Ollama); a per-model block-list catches models not covered by regional processing.
Operational autonomy & third-country access protection
No provider operator, hyperscaler or third-country authority may gain access or compel disclosure without your authorisation — the system must allow no silent operator override (CLOUD Act / FISA 702, BSI 'Cyber Dominance').
Fail-closed residency gate enforced at preflight (HTTP 409 before any spend) and at runtime (`DataResidencyViolation`) — with no operator bypass: `?preflight=skip` cannot override residency FATALs, and every skip is audited. The dedicated host offers no provider-side control plane to subpoena; the fully-EU model path removes any CLOUD Act nexus for inference. Chinese-jurisdiction models are hard-blocked under `eu`.
Exit & portability (EU Data Act Ch. VI)
Full data and workflow export plus a contractually and technically supported exit path — no vendor lock-in (EU Data Act switching and portability rights).
Workflows are declarative, human-readable YAML DAG templates — portable, inspectable artifacts the customer holds, not opaque vendor config. Data lives in open, exportable stores (PostgreSQL/Apache-AGE, Qdrant, Valkey, SeaweedFS); artifacts are addressable via stable `/api/v1/artifacts/<wf>/<label>` JSON endpoints. Multi-provider routing (LiteLLM) prevents model lock-in; the whole platform is the customer's dedicated deployment (git pull + docker compose).
Transparency & auditability
Every AI action, model call, cost and policy decision must be logged as durable evidence and be queryable — suitable for EU AI Act record-keeping (Art. 12) and GDPR accountability (Art. 5(2)).
A SQL-queryable `notification_log` records every governance event with structured categories (residency block, runtime violation, skip veto, MCP connector). A cost ledger writes every LLM call to `llm_cost_events` (model, cost in EUR, latency); Temporal provides full event replay; Prometheus counters expose enforcement metrics. Langfuse, the Temporal UI and Grafana are pre-wired.
Model & supply-chain control
The buyer must control which models and providers process their data — with a documented EU allow-list and a path to a fully-EU / self-hosted supply chain free of third-country dependency (BSI C3A, NIS2 supplier risk).
Multi-LLM routing across multiple providers with an enforced EU allow-list under `eu` and a fully-EU path (Mistral FR managed + self-hosted Ollama on the customer host). New model IDs are verified against official provider docs and registered before use — no hallucinated endpoints. A per-model block-list catches models not covered by regional processing; managed failover keeps availability without violating residency.
Defence against AI-specific attacks
The system must detect and flag AI-specific threats — prompt injection, jailbreaks, output-side data exfiltration, tool abuse, package hallucination (OWASP GenAI Top-10, EU AI Act Art. 15 robustness, BSI).
PromptGuard, a defence-in-depth architecture (seven layers live), detects and flags via a typed middleware: an input scanner, an injection classifier, an output rogue-string detector, a package-hallucination validator (PyPI/npm check) and an egress sanitiser, plus memory-write quarantine and SHA256 integrity checks of MCP tool descriptions. A continuous Garak red-team CI tests effectiveness. Honest framing: several phases currently run in observe-only mode (detect/flag); the metrics come from internal benchmarks.

Why this holds up under qualification scrutiny

Single-tenant, not shared SaaS

The full stack runs per customer on a dedicated EU host — its own Postgres, its own vector and cache layers, its own object store. There is no shared data plane for a third country to subpoena or an operator to silently override.

Fail-closed residency — no operator bypass

Violations fail at preflight (HTTP 409, before any spend) and at runtime. `?preflight=skip` technically cannot bypass residency FATALs, and every skip attempt is audited. The fully-EU model path (Mistral FR + Ollama) removes any US CLOUD Act nexus for inference — provable by audit log.

Audit ledger doubling as AI Act Art. 12 evidence

Every LLM call produces a cost row in EUR, every governance event a structured row in `notification_log`, every workflow a full Temporal event replay. That delivers the end-to-end, SQL-queryable evidence trail that record-keeping duties and GDPR accountability demand.

Sovereign AI is now German procurement reality

Initiatives such as ZenDiS with its open-source suite openDesk, the German public-sector cloud strategy and Gaia-X are driving the market. It is also reported that a German Telekom/SAP consortium was first-ranked for a sovereign federal AI platform (Deutschland-Stack, reportedly around €250M). AIonicOS is aligned with exactly this sovereignty trajectory — as a platform you control yourself.

Frequently asked questions from procurement and IT security

Where are my data processed and stored?

AIonicOS runs exclusively in certified data centres within the European Union. Your tenant operates on dedicated resources (single-tenant), and data never leaves the agreed EU legal jurisdiction – including for backups or telemetry.

Does startvisor.AI or any third party have access to our content and model inputs?

No. In single-tenant mode, model inference and data pipelines are fully tenant-isolated. startvisor.AI operations staff receive only time-limited, logged break-glass access for incident response – never routinely and never for AI training.

How do we ensure we can switch providers or migrate off the platform (exit strategy)?

AIonicOS exports all workflow definitions, model artefacts, and operational data in open standard formats (YAML, JSON, OCI images). A documented exit plan with defined data-portability SLAs is part of every enterprise contract, addressing the EU Data Act's switching and portability rights requirements.

How does AIonicOS support EU AI Act requirements – particularly for high-risk applications?

AIonicOS is designed to align with EU AI Act requirements: embedded human-in-the-loop design, immutable audit logs, explainability modules, and risk-classification workflows. We do not certify conformity for your specific application – that responsibility lies with you as the deployer – but we provide the technical prerequisites for structured compliance documentation.

What certifications and evidence can startvisor.AI provide for a procurement procedure?

Our operational infrastructure uses ISO-27001-certified data centres. AIonicOS itself is on a path toward its own certification; current evidence includes penetration test reports, GDPR processing records, data processing agreement templates, and technical security documentation. For formal procurement procedures we provide a complete qualification evidence package on request.

Book a sovereignty and procurement-readiness call

In 30 minutes we walk through your procurement and sovereignty criteria — data residency, third-country exposure, the EU Data Act exit path and the evidence your qualification process demands. You leave with a clear mechanism map, not slides.

AIonicOS is compliance-enabling and aligned with recognised frameworks (including BSI C5, the BSI C3A criteria, the EU Cloud Sovereignty Framework, the EU Data Act and the EU AI Act) — this is neither a certification nor legal advice. The BSI C3A criteria are a voluntary assessment framework, not a certificate; AIonicOS is not represented as 'C3A certified'. Regulatory classification of your specific application remains with you as the controller; startvisor.AI acts as an Art. 28 GDPR processor and is not a law firm. Any cost figures are estimates.