Sovereign AI for National Government
Ministries and agencies hold the data a state is made of, some of it classified, all of it under a legal basis that names one controller. AI at national scale has to respect both, and it has to run on infrastructure the state controls.
Air-gapped by design
The complete platform runs on classified networks with no outbound connection. Updates arrive through the accredited channel. Clearance is enforced at retrieval.
Controllers stay controllers
Ministries publish governed data products and analysts join them for an authorised question, under a named legal basis, without a central copy.
One stack to accredit
The same platform, governance and audit on the unclassified and the classified side, which is what the security authority and the EU AI Act auditor want to see.
A national government is a federation of controllers, each with a legal basis for its own data and none for the others', and a growing share of that data sits on networks that must never touch the internet. Ministries have run AI pilots on unclassified copies and hit the wall on both counts. The platform gives a ministry, an agency or a cross-government service AI on its own data, in its own perimeter, including disconnected, with the governance a state auditor and the EU AI Act require.
Read this if you're leading digital transformation, IT, security or data at a ministry, a federal or national agency, a statistical office or a national security body, or accountable for its EU AI Act, GDPR and NIS2 obligations.
For national and federal government, Scrydon is the sovereign AI and data platform that runs models, agents and analytics on ministerial and agency data inside state-controlled infrastructure, including fully air-gapped classified networks, and lets ministries share governed data products without any of them giving up control.
It runs identically in a national sovereign cloud, on government data centres and on disconnected classified networks, so there is one platform to accredit and one governance model from the unclassified service desk to the classified enclave.
Data that cannot be pooled, networks that cannot be connected
A national government is a federation of controllers. The tax administration, the social security institution, the housing agency and the health ministry each have a legal basis for their own data and none for the others', so the traditional answer to a cross-cutting policy question, a central data warehouse, either never gets built or gets built without the data that matters. Meanwhile a growing share of the state's most important data sits at RESTRICTED and above, on networks that never touch the internet, where there has been no AI at all, or shadow use on unclassified copies.
Every ministry has by now run a pilot and met the same two walls: the data cannot be pooled, and the network cannot be connected. Behind those stand the security authority that has to accredit whatever runs on classified data, the EU AI Act that asks the state to know and govern every AI system it uses, and a court of audit and a parliament that expect the decision to be explainable. Strategic autonomy is not a slogan here; it is the operating condition.
Separate controllers by law — Tax, social security, housing and health each have a legal basis for their own data. A central warehouse is either never built or built without the data that matters.
Classified networks with no AI — RESTRICTED and above never touch the internet, so those networks have had no AI at all, or shadow use on unclassified copies.
Shadow AI across ministries — Staff use consumer tools because nothing sanctioned exists. Discovering and governing that is an EU AI Act readiness task in itself.
Accreditation and audit — Every deployment has to be assessed by the security authority and explained to the court of audit and parliament.
Air-gapped where it must be, federated where it can be
The platform is deployed inside the perimeter that applies: connected in a national sovereign cloud or a government data centre for the unclassified estate, and air-gapped on the classified network, with the same stack in both. On the classified side, open-weight models, retrieval indexes and agents run entirely inside the enclave, updates arrive through the accredited import channel and are recorded, and document clearance markings are enforced at retrieval time so nothing is served above the reader's clearance.
Across ministries, a data space reverses the direction of the warehouse: each controller keeps its data where it is and publishes governed data products whose schema, quality and access policy are described in a shared ontology. A cross-ministry question is a query over those products, executed under an authorisation that names the legal basis, the purpose and the retention period, and logged. Where the law allows only aggregate or pseudonymised results, the product enforces it. Governance, agent identity and the model inventory then produce the evidence the security authority, the EU AI Act and the auditor ask for.
- 1
Deploy in the perimeter
Connected in a national sovereign cloud or government data centre; disconnected on the classified network, with the same stack.
- 2
Ground in the ministry's own sources
Open-weight models and retrieval run on the ministry's documents and records, with clearance markings enforced at retrieval time.
- 3
Federate across ministries
Each controller publishes governed data products into a data space; a cross-ministry question runs under a named legal basis and is logged.
- 4
Govern and evidence
Policy, identity, audit and the model inventory produce the evidence for the security authority, the EU AI Act and the auditor.
Strategic autonomy is an operating condition, not a slogan
The platform is designed so that no foreign dependency sits in the loop of a national deployment. Models are open-weight and served on state-controlled infrastructure; there is no inference call, licence check or update path outside the perimeter, and the disconnected case is the design case rather than a degraded mode. That is what lets one platform serve the service desk, the policy unit and the classified enclave, and what keeps accreditation to one assessment and one evidence set.
Control is enforced where the data is used. Clearance is metadata on every document and is applied at retrieval, so an analyst with a lower clearance never sees a passage above it, even inside the same network. A cross-ministry query carries its legal basis and is logged against it. Agents act under scoped identities with permissions per system and per action, and every action is attributed. The audit log, policy decisions and model inventory that result are the EU AI Act, GDPR and NIS2 evidence, produced by running the platform, not written for the auditor afterwards.
No foreign dependency in the loop — Open-weight models served on state-controlled infrastructure. No inference call, no licence check and no update path outside the perimeter.
Clearance at retrieval — Document classification is first-class metadata. An analyst with a lower clearance never sees a passage above it, even inside the same network.
Legal basis in the query — A cross-ministry query carries the legal basis, the purpose and the retention period, and the data product enforces aggregate or pseudonymised results where the law requires.
Evidence, not a slide — The audit log, policy decisions and model inventory are the accreditation evidence, produced by running the platform, not written afterwards.
Keep an eye on this space
What changed for your sector in European AI sovereignty, and what we learned in the field. A few times a year, no drip campaign.
Use cases for national government
Classified networks, cross-ministry analytics and governed AI at state scale.
Back Office Automation
Administration
Challenge
Manual approval processes, procurement delays, and administrative burdens slow down operational readiness and consume valuable personnel hours.
Solution
Automate common approval use cases, procurement requests, and HR processes with intelligent agents that ensure policy compliance.
Read more
Streamlined operations allowing personnel to focus on mission-critical tasks rather than paperwork.
Cross-Agency Benefit Adjudication
Social Services
Challenge
Citizen applications require verification across Tax, Health, and ID agencies, but sharing raw data violates privacy statutes.
Solution
Federated agents query data products across agency domains, returning only 'Yes/No' eligibility proofs based on pre-approved policy code.
Read more
Instant adjudication without creating a central database of citizen data, fully compliant with GDPR and local privacy laws.
Citizen-Developer Case Automation
Digital Services
Challenge
Departments have a backlog of small, repetitive workflows — a permit reminder here, an intake form there — but IT teams cannot build a bespoke agent for every one.
Solution
A governed no-code builder lets policy and operations staff assemble their own AI agents from approved building blocks, with every workflow still subject to the same access controls and audit trail as an engineered agent.
Read more
Dozens of small workflows automated by the teams that own them, without adding to the central engineering backlog or bypassing governance.
Shadow AI Discovery for EU AI Act Readiness
Compliance & Risk
Challenge
Individual teams have quietly adopted their own AI tools and agents, so the organisation cannot produce a complete inventory ahead of EU AI Act high-risk obligations.
Solution
Continuous discovery surfaces every agent and AI tool in use across departments, classifying each against EU AI Act risk categories and routing unsanctioned use into governed alternatives.
Read more
A complete, defensible AI inventory and a measurable reduction in unauthorised tool use ahead of the 2 December 2027 enforcement deadline.
Human-in-the-Loop Policy Exception Handling
Case Escalation
Challenge
Most applications follow a straightforward rule, but the exceptions — a hardship case, an unusual combination of circumstances — need a human decision, and today those cases get lost in the same queue as routine ones with no clear handoff.
Solution
The AI OS routes the routine majority through automated processing while recognising genuine exceptions and handing them to the right official with full context already assembled, then carries the human's decision back into the process automatically.
Read more
Exceptions reach a human reviewer faster and with better context, while the routine majority never needs one at all.
One Citizen Record Across Departments
Master Data
Challenge
Tax, health and licensing systems each hold their own version of "this citizen" or "this business," with different identifiers and slightly different details, so cross-department work means guessing which records actually refer to the same person.
Solution
An ontology defines each citizen and business as a single entity, resolved once across every source system, so departments and agents reason over one consistent identity instead of reconciling mismatched records.
Read more
Cross-department processes start from one agreed identity instead of a manual matching exercise, cutting errors and delay.
What National & Federal Government has to comply with
The regulations that decide whether AI can run on this data at all. Each page lists what the framework asks and which platform controls answer it — properties and refusals, not a checklist.
AI Act · EU AI Act
Applies to: Providers, deployers, importers and distributors placing AI systems on the EU market or whose AI output is used in the EU.
How the platform supports itGDPR · General Data Protection Regulation
Applies to: Any organisation that processes the personal data of people in the EU/EEA, whether established in the Union or offering goods, services or monitoring from outside it.
How the platform supports itNIS2 · NIS2 Directive
Applies to: Essential and important entities across critical sectors — energy, transport, water, health, digital infrastructure, public administration, manufacturing and more — and their supply chains.
How the platform supports itSecNumCloud
Applies to: Cloud service providers seeking French ANSSI qualification, and the public-sector and sensitive-data organisations that require qualified, sovereign cloud services.
How the platform supports it
What these outcomes actually run on
A ministry's architect asks whether it runs disconnected, how AI use is governed and how ministries share. These are the pages that answer those questions.
Sovereign Foundations
The zero-trust foundation the whole platform sits on — the same stack from air-gapped to cloud.
Air-Gapped
The full platform running offline, on networks with no route to the internet.
AI OS
The runtime that maps a process to steps, routes each step to a system, an agent or a person, and gives it the context to act.
AI Governance
Policy, data-loss prevention and audit applied to every AI action, with the evidence an assessor can read.
Data Spaces
Data shared across organisational boundaries without handing over ownership of it.
Identity
Federated identity and zero-trust access — for people and for agents, under the same policy.
Frequently asked questions
How do air-gapped AI deployments work for sensitive government data?+
Can ministries analyse data together without a central data warehouse?+
How does the platform help with EU AI Act readiness across a government?+
Is one platform enough for both the unclassified and the classified side?+
What about a national sovereign cloud or a government data centre?+
How is agent access to ministerial systems controlled?+
Do you partner on tenders and framework contracts?+
Prefer to write? Email hello [at] scrydon.com and we will get back to you.
Keep going
Three routes from here: the rules you will be measured against, the platform these outcomes run on, and the sessions where we walk through them.
The rules that apply
Each regulation, what it asks, and the controls the platform provides for it.
The platform behind it
The AI Operating System, Analytics and Sovereign Foundations, page by page.
Every use case
All of them in one place, grouped by the sector that knows them best.
How – Inside the AI OS: running governed agents on your own cluster, live
17 Sept 2026, 09:00
Part 2 of the Sovereign AI series, for engineers and architects. No slides after minute five: an agent gets an identity and scoped permissions, calls tools over governed MCP, retrieves from the ontology rather than raw tables, runs inside the sandbox and is stopped when it steps outside policy, and everything lands in the audit trail — then the same stack brought up on a disconnected network. Properties and refusals, shown rather than claimed.
Other parts of Government
The other dedicated pages under Government, and the sector overview they hang off.
EU Institutions, Bodies & Agencies
Institutions, bodies and agencies of the Union: multilingual AI on European infrastructure, federated data with member states, and governance to the standard of the AI Act.
Read the pageLocal & Municipal Government
Cities, municipalities and regions: multilingual citizen assistants, permit and benefit case work, and public-works analytics on the municipality's own rules and data.
Read the pagePolice & First Responders
Police, fire, dispatch and civil protection: case-file analysis, dispatch support and multi-agency incident pictures, inside the service and admissible.
Read the page
Partners
