Sovereign AI · Roadmap
A sovereign AI platform for European organisations.
European institutions and organisations should have a European alternative for managing their data, documents and applications, with AI that works under their control. SovereignEU is developing this alternative with a clear ambition: developed in Europe, hosted in Europe and operated by European organisations.
In development. Availability, compatibility and migration paths will be documented for each release.
SovereignEU · Platform concept
Your applications, connected.
- Documents Planned alternative to SharePoint
- Knowledge Planned alternative to Confluence
- Office Planned alternative to Microsoft 365 (Word, Excel and PowerPoint)
- Intranet Planned alternative to SharePoint · Viva
One point of access
Your AI workspace
What would you like to do?
Find information. Create documents. Coordinate work.
What the alliance brings
One alliance, three roles inside your organisation.
SovereignEU is a European public and private alliance. It advances research, develops a planned sovereign operating environment, and helps your organisation assess and adopt it responsibly.
Open research & development
Sovereign AI advanced in the open, with European institutions and industry, so the foundations stay inspectable and shared.
ExploreSovereignOS & applications
A planned workstation operating system and a suite of organisational applications, designed around customer control.
ExploreProfessional services
A measured path from one bounded use case to a validated deployment: assess, define, deploy, then verify and transfer.
ExplorePlatform architecture · Roadmap
One platform. Four connected layers.
The planned architecture connects your data, approved AI and everyday applications, with one natural-language interface to bring them together.
-
Interface layer
Natural-language access
Ask questions. Find information. Request actions.
One conversational entry point across the applications your role permits, with human approval where required.
Explore the AI workspace -
Application layer
The tools for daily work
-
Intelligence layer
AI models & agents
Approved models and agents, governed by your deployment policy.
- Local modelsRun within an approved deployment.
-
Sovereign provider modelsProviders assessed for jurisdiction and data handling.
- AI GuardrailsApply the organisation’s rules.
- AI agentsUse approved tools for bounded tasks.
-
Unified data layer
Sovereign DB
A planned unified foundation combining graph data, vector storage, time series, a triple store and full text search, with hybrid search across these capabilities.
- Graph
- Vector store
- Time series
- Triple store
- Full text search
- Hybrid search
Controls across every layerIdentity, permissions, data boundaries, approvals and reviewable actions.
Concept architecture · Roadmap. Data formats, search capabilities, model providers and application integrations require validation for each release.
The product surface · Roadmap
Working inside SovereignOS.
Pick an example task. SovereignOS keeps the work inside your data boundary, routes it to an approved local model, drafts with citations, and waits for a person before anything is shared.
Prepare the mobility committee brief from approved project records.
Decision brief: shared mobility procurement
A cited draft assembled only from records the organisation has approved, ready for a named official to review before anything leaves the building.
6 approved sources · 2 passages need review
Draft a reply to the accessibility enquiry, citing current policy.
Reply: accessibility of the mobility portal
A plain language reply grounded in the accessibility standard and the current service policy, with each commitment linked to its source.
4 approved sources · 1 passage needs review
Summarise this quarter's programme spend for the board.
Summary: Q3 programme spend
A board-ready summary built from the finance system of record, figures reconciled, variances flagged, nothing invented.
5 approved sources · figures reconciled to ledger
Roadmap interface concept, illustrative, not a released product.
How the platform is organised
Eighteen roadmap capabilities, four clear responsibilities.
Everything in SovereignOS belongs to one of four responsibilities, so an official always knows what a capability is for, and an evaluator knows where it sits.
Applications
The organisational apps people use every day, documents, knowledge, service desk, projects and more.
ExploreIntelligence
Document AI, assistants and fine-tuned models, governed, cited, and kept on approved routes.
ExploreFoundation
The AI native database, model runtime and identity layer the whole system stands on.
ExploreTrust
Guardrails, AI defense and post quantum cryptography, the controls that make the rest safe to use.
ExploreSovereign Intelligence
How a governed answer is actually produced.
Five steps, each one accountable. Select a step to see what happens, and what stays under your control.
01
Connect
02
Retrieve
03
Reason
04
Act
05
Evidence
Roadmap interface concept, illustrative, not a released product.
Solutions by function
It meets each part of the organisation where it works.
The same governed core, shaped to the daily reality of programme teams, operations, IT and people functions.
Programme delivery
Keep delivery, decisions and evidence in one place, briefs drafted from the record, not from memory.
Operations & administration
The everyday apps and documents, with the boundary and the audit trail built in.
IT & service management
Requests, changes and service records, governed, traceable, and yours to run.
HR & people
Sensitive people processes handled inside your control, with human approval where it matters.
Governed by design
European governance built into the plan.
SovereignOS is designed around four obligations European officials already answer to.
Data protection
Designed to support GDPR obligations with a defined data boundary, lawful basis and retention policy for each deployment.
AI lifecycle governance
Plan how AI will be used, assess its risks and keep a record of changes.
Human oversight
A person can be required to approve sharing or consequential actions. The gate is part of the planned design and can be defined for the deployment.
Technical evidence
A reviewable activity record gives your organisation evidence to assess its obligations.
What “sovereign” actually means here
Six dimensions you can verify, not a slogan.
We hold ourselves to a definition an evaluator can check, across every dimension of control.
Data
Designed to let your organisation choose where data is processed and how approved models may use it.
Operational
You can run, pause and audit the system with operator access defined by your organisation.
Technology
Planned open source releases intended for inspection, adaptation and an orderly exit under their published licences.
Jurisdiction
Designed for deployments where jurisdiction, governance and access routes are reviewed and documented.
Knowledge
Your organisation can retain institutional memory under its own policy, with export and access rules defined up front.
Exit
A credible way out: your data and models are portable, so sovereignty is real, not rhetorical.
A measured path
Start with one bounded use case. Prove it. Then widen.
No attempt to transform everything at once. We begin where the value and the risk are both legible, and we transfer control to your team at the end.
01
Assess
02
Define
03
Deploy
04
Verify & transfer
Common questions
The questions we hear first.
Straight answers for nontechnical leaders, why a new OS, who builds it, your data, cost, and how to start.
Start a conversation
Bring one bounded use case. We’ll show you the governed path to a validated deployment.
Public-sector teams across the union are planning for sovereign AI. A first deployment starts with a single, well-chosen decision.