National Programmes: Who Is Buying Ontology, Country by Country
Ontology work is bought by governments far more often than by companies, and the budgets are public if you know where to look. Private sector demand is real but scattered, usually hidden inside a data platform programme, and rarely named as ontology. Public sector demand is named, funded, and published in procurement notices with the standard written into the specification. This lesson maps the programmes that are actually spending, what each one has standardised on, and what that means if you are choosing where to point a career or a bid.
United Kingdom: 4D by policy
The National Digital Twin Programme sits in the Department for Business and Trade and is the anchor buyer. It funds and leads an open source data sharing infrastructure called the Integration Architecture, with code released under Apache 2.0 and documentation under the Open Government Licence, and it has made the Information Exchange Standard the common ontology of that architecture. IES is a 4D ontology developed inside government with Dstl, the Ministry of Defence, the Metropolitan Police Service, the Foreign, Commonwealth and Development Office, the Home Office, HMRC and DBT as joint custodians, published openly with MIT-licensed code. The programme is extending it with IES:Building, a 4D extension for the built environment, and has gone to market for ongoing ontologist support to extend and refine the standard. When a government publishes a contract notice for ontologists, the demand signal is no longer theoretical. Programme information is at ndtp.co.uk.
The rest of the UK picture reinforces it. GCHQ publishes HQDM and MagmaCore under Apache 2.0 with Crown Copyright, giving the 4D community a maintained RDF implementation. The National Underground Asset Register, now operated by Ordnance Survey, was put on a statutory footing by the Data (Use and Access) Act 2025, which received Royal Assent in June 2025, and is projected to deliver around £490 million a year in economic benefit through fewer asset strikes and less disruption. NHS England has mandated SNOMED CT since April 2018 in primary care and since April 2020 across secondary, acute, mental health, community, dentistry and optometry systems. The UK is, in practice, the most 4D-committed jurisdiction in the world, and that creates a peculiar local labour market: people who understand states, temporal parts and extensional identity are scarce and are being hired.
United States: BFO as baseline, and a public knowledge network
The American picture divides into defence and science. On the defence side, in 2024 the chief data officers of the Department of Defense, the Office of the Director of National Intelligence and the Chief Digital and Artificial Intelligence Office adopted Basic Formal Ontology and the Common Core Ontologies as baseline standards for formal ontology work. BFO is standardised as ISO/IEC 21838-2:2021; CCO is an eleven-module mid-level suite originally created by CUBRC with IARPA funding and available under a BSD-3 licence. That decision propagates into every downstream data programme specification and is the single most consequential procurement event in the field’s history.
On the science side, the National Science Foundation invested $26.7 million in eighteen projects under the Prototype Open Knowledge Network programme, run in collaboration with NASA, the National Institutes of Health, the National Institute of Justice, NOAA and the US Geological Survey. Fifteen projects build knowledge graphs against societal challenges, two build the interconnecting fabric, and one covers education and public engagement. The fabric project, FRINK, led from the University of North Carolina at Chapel Hill with RENCI, provides shared access across the resulting graphs, and there is now a Model Context Protocol server exposing more than thirty of these scientific knowledge graphs with graph routing, schema inspection and ontology-driven query expansion, described in arXiv 2605.30283. Programme details are at proto-okn.net. NIST contributes on the manufacturing side, publishing analysis of the Industrial Ontology Foundry Core ontology and of the ISO 23247 digital twin framework series.
European Union: catalogues, data spaces and profiles
The European approach is less about upper ontologies and more about interoperable metadata at scale. The SEMIC action of Interoperable Europe maintains DCAT-AP, the application profile for describing datasets and data services in European data portals. DCAT-AP 3.0 reached SEMIC Recommendation status and was published on 14 June 2024, aligned with W3C DCAT 3, adding dataset series and applicable legislation among other changes. It is the profile that data.europa.eu and national portals implement, and there are derived profiles such as the high value dataset profile. SEMIC has also published DCAT-AP for Data Spaces material, which is the connective tissue between the catalogue tradition and the newer common European data spaces created under the Data Governance Act and Data Act framework. The specification is public at the DCAT-AP 3.0 release site.
Norway deserves separate mention as the industrial ontology powerhouse. The READI joint industry project, running from 2018 to 2022 with Equinor, Aker BP, Aibel and Aker Solutions, produced a digital reference data library aligned to what is now the Industrial Data Ontology in the ISO 23726 series, tied to real field development projects. Nordic road and rail authorities agreed in 2017 to require openBIM delivery, which is why IFC adoption in infrastructure is further ahead there than in most of Europe.
International bodies
Three multilateral organisations run vocabularies that function as global public infrastructure. The Food and Agriculture Organization maintains AGROVOC, whose July 2025 release carries 41,447 concepts and 1,219,844 terms in up to 42 languages, published as SKOS linked open data with a SPARQL endpoint. SNOMED International governs SNOMED CT on behalf of 53 member countries, with free use in member territories and annual affiliate licensing elsewhere. The World Health Organization maintains the International Classification of Diseases, whose eleventh revision was adopted by the World Health Assembly in 2019 and came into effect on 1 January 2022, and which is now natively digital with an API rather than a book. Between them these three shape health, agriculture and food data policy in most countries on earth.
| Jurisdiction or body | Programme | Standard adopted | Access reality | What they buy |
| United Kingdom | National Digital Twin Programme, DBT | Information Exchange Standard, 4D, with IES:Building extension | Open GitHub, MIT code and Open Government Licence documentation; Integration Architecture under Apache 2.0 | Ontologist support, integration architecture engineering, domain extensions |
| United Kingdom | NHS England clinical coding | SNOMED CT | Free in member territory; affiliate licensing outside | Terminology services, mapping from legacy code systems, tooling |
| United States | DoD, ODNI and CDAO data offices | BFO with Common Core Ontologies as baseline from 2024 | ISO/IEC 21838-2:2021 for BFO; CCO under BSD-3 on public GitHub | Mission ontologies conformant to the baseline, alignment and validation work |
| United States | NSF Proto-OKN with five partner agencies | Open knowledge graphs plus a shared access fabric | $26.7 million across eighteen projects; public graphs with MCP access | Domain knowledge graph construction, interconnection infrastructure, outreach |
| European Union | SEMIC, Interoperable Europe | DCAT-AP 3.0, published 14 June 2024, aligned to DCAT 3 | Open specification and profiles; implemented by data.europa.eu and national portals | Catalogue profile implementation, data space interoperability, harvesting pipelines |
| Multilateral | FAO, SNOMED International, WHO | AGROVOC, SNOMED CT, ICD-11 | AGROVOC open with SPARQL; SNOMED members-only free tier; ICD-11 in force since 1 January 2022 | National implementation, translation, mapping and terminology services |
What breaks · mandating a standard without conformance criteria
A programme writes the sentence that appears in procurement documents everywhere: all suppliers shall deliver data conformant to the mandated standard. No conformance clause is cited, no test suite ships, and no gate runs at ingest.
Three suppliers each deliver something defensible: one maps its existing warehouse extract to the standard’s class names, one delivers a full model with every optional property empty, one delivers a rich model under its own identifier scheme. All three are conformant under the only reading available, because using the terms of a standard is all the mandate actually required, and none of the three can be joined to the others. The programme has bought vocabulary alignment and paid for interoperability. The reading move the standards lesson teaches is what prevents this: find the conformance section first, ask whether an automated checker could decide it, and measure the gap between the conformance bar and the interoperability bar before signing. The fix is unglamorous and cheap by comparison: publish the shapes, the test fixtures and the gate alongside the mandate, so conformance becomes something a build either passes or fails.
Check yourself
A national programme wants its mandate to produce genuinely interoperable deliveries. Which requirement does the most work?
- Published validation artefacts, shapes and fixtures, run as a gate at ingest, with results returned to the supplier.
- A requirement that suppliers use the standard’s class and property names.
- A requirement that suppliers hold a certification from a recognised body.
- A requirement that every delivery be accompanied by documentation of the mapping.
Show the answer, and why each wrong option is wrong
A is correct. An executable gate converts a mandate into a decidable question, gives suppliers red-and-green feedback instead of email archaeology, and makes non-conformance visible before the data is in the estate. It is the same shapes-as-contract move the register used on its three source systems.
B is the mandate that fails above: name usage permits three mutually unjoinable deliveries. C certifies the organisation rather than the payload, and organisations with excellent certifications ship bad files under deadline pressure like everyone else. D is genuinely valuable and still not sufficient: documentation describes the intent of a mapping, while only a gate tests what actually arrived, and the gap between the two is where every integration incident lives.
What this means for careers and bids
Three practical conclusions. First, the demand is public sector led and specification driven, which means the winning credential is not a certificate but a public artefact: a crosswalk, a validated ontology, a SHACL suite that somebody else can run. Buyers in this space read repositories. Second, the jurisdictions have diverged on the deepest modelling question, with the UK committing to 4D through IES and HQDM while the US committed to BFO’s 3D continuant and occurrent split. Anyone working across allied programmes will therefore be doing alignment work, which is why the crosswalk skill set has a higher market value than authoring skill alone.
Third, licensing and sovereignty shape the tooling. Defence and health environments frequently cannot call a hosted service, so standards with permissive licences and offline implementations, IES, HQDM, BFO, CCO, GeoSPARQL, OBO ontologies, win by default over anything requiring a vendor relationship. If you are positioning for this work, build in the open, publish the tests, and make your artefacts runnable inside an air-gapped enclave. That is the profile these programmes are buying, and the specialist tracks in our ontology training collection follow the same domains as the procurement pipelines described here.
