Why transforming beneficial ownership data into linked data formats matters for AI — and for tackling financial crime
International beneficial ownership data is inherently complex.
A single corporate structure can span dozens of entities across multiple jurisdictions, with control exercised through chains of direct and indirect shareholdings, voting rights, nominee arrangements, and trust structures.
For a human investigator, tracing that chain is already hard work. For an AI system trying to reason over it, the challenge is even more fundamental: how do you make the machine understand what the data means?
That is exactly the problem that the Beneficial Ownership Data Standard (BODS) RDF Vocabulary addresses and why it deserves more attention from both the knowledge graph community and the growing number of teams applying AI to financial crime, sanctions compliance and corporate transparency.
From raw data to machine-readable meaning
My previous posts have looked at how BODS helps governments collect and share corporate ownership information in a consistent format, and how standardisation underpins the value of that data.
But standardising the schema is only part of the story. A machine receiving a BODS dataset still needs to understand the semantics of what it is looking at — that a personStatement represents a natural person, that relationshipStatement describes a directed relationship between two entities, that interests can carry percentage thresholds and types of control.
The BODS RDF Vocabulary shows how to map the full BODS 0.4 data model — persons, entities and ownership-or-control relationships — to a formal ontology. It also includes information about when and why the RDF data model deviates from the JSON representation of BODS.
The formal ontology structure, expressed in RDF and Turtle, does something crucial: it makes the semantics of beneficial ownership machine-interpretable, not just machine-readable.
The knowledge graph community has understood for some time that ontologies are what make graph data analytically powerful.
Work from Graphwise on ontologies and knowledge graphs in the financial industry illustrates this clearly: an ontology-driven approach enables automated reasoning — detecting hidden entity relationships, flagging suspicious transaction patterns, supporting KYC and AML workflows — in ways that simple structured databases cannot.
The BODS RDF Vocabulary brings exactly that capability to beneficial ownership data.
What this means for AI systems
The rapid development of large language models and graph-native AI applications has made the question of semantic data structure more important, not less.
LLMs struggle with complex relational data unless it is presented in a form they can parse meaningfully. A raw BODS JSON array, without context, offers little for a model to work with.
But BODS expressed as an RDF knowledge graph, grounded in a formal ontology, becomes something quite different: a traversable, queryable structure that can express concepts like “find all entities where the beneficial owner is also a PEP” or “identify the ultimate parent of this company across an indirect chain” in SPARQL, and increasingly, in natural language queries translated to SPARQL by an LLM.
The BODS risk detection proof-of-concept I worked on in 2023 demonstrated exactly this kind of cross-dataset reasoning — combining BODS RDF data from the UK with global sanctions data and national public procurement records to surface hidden risks in ownership networks.
The SPARQL examples published alongside the vocabulary show how queries for UBOs, ultimate parents and ownership or control chains are expressible in a handful of lines once the data is in RDF.
It is also worth noting that the BODS RDF Vocabulary is influencing the work of others. The Senzing team — who have one of the leading entity resolution platforms — have published sz-semantics, an RDF Metadata Application Profile for representing entity resolution results as a knowledge graph.
It explicitly aligns with BODS alongside other vocabularies such as NIEM Open and FollowTheMoney, reflecting that serious practitioners in the identity resolution space are treating BODS as a credible and stable reference vocabulary.
When an entity resolution platform chooses to align its RDF namespace with BODS, it signals something important: that the standard has achieved sufficient maturity and adoption to serve as part of broader AI and data integration infrastructure.
The standardisation argument
None of this works without standardisation. An RDF vocabulary built on inconsistently collected, jurisdiction-specific beneficial ownership data is still an unreliable foundation for AI-driven analysis.
The value of the BODS RDF Vocabulary is inseparable from the value of BODS itself: a common data model that works wherever the underlying data comes from.
Last year’s update to the BODS RDF Vocabulary aligned it with BODS version 0.4, bringing in the temporal change-tracking and indirect chain representation that makes real-world ownership data understandable and usable.
That alignment between the standard and its linked data expression is what makes it possible to build AI applications that reason reliably across diverse, cross-jurisdictional datasets.
For practitioners building financial crime detection tools, compliance platforms, or investigative data infrastructure, the BODS RDF Vocabulary is worth taking seriously.
For four years, I was the product owner of the Beneficial Ownership Data Standard, the world’s leading open standard for beneficial ownership information and continue to advocate to improve beneficial ownership data quality and standardisation. To learn more or if you need technical guidance or consultancy support, contact me via beneficialownership.co.uk.
