What is machine-readable regulation?
Machine-readable regulation is legislation, regulation, standards or policy represented in a structured format that software can parse reliably: documents, articles and provisions with stable identifiers, defined concepts, cross-references and metadata such as dates and jurisdiction. It makes regulatory text navigable and linkable by machines. It does not by itself make the rules executable; that is the next step, machine-executable regulation.
What machine-readable regulation contains
A useful machine-readable representation goes well beyond a PDF with text extraction.
- Document structure: titles, chapters, articles, paragraphs and points, each addressable by a persistent identifier.
- Provisions and their relationships: references, amendments, repeals and dependencies between rules.
- Concepts and definitions: terms the regulation defines and where they apply.
- Requirements: obligations, permissions, prohibitions, conditions and exceptions, tagged to the provisions that create them.
- Metadata: versions, dates of application, jurisdiction, authority and language.
Standards for machine-readable regulation
Akoma Ntoso, an OASIS standard, structures legislative and legal documents. LegalRuleML, also from OASIS, represents the rules those documents contain. prEN 18286 and related European work aim at interoperable digital regulatory information. Knowledge graphs and Semantic Web technologies connect all of these into regulatory knowledge that can be queried.
Machine-readable versus machine-executable
Machine-readable regulation lets software find and link the right provision. Machine-executable regulation lets software evaluate whether a situation satisfies the rule. The first is about structure; the second is about formal rules with inputs, outputs and deterministic evaluation. NormLogic produces both, and keeps the executable rule connected to the readable source it came from.
How NormLogic produces machine-readable regulation
NormLogic ingests source documents, identifies structure, provisions, concepts and requirements with AI assistance, and records each element with provenance to its source. The result is a regulatory knowledge graph that applications, compliance checks and AI agents can use, and that can be exported in interoperable formats.
Frequently asked questions
Is a PDF of a regulation machine-readable?
Only in the weakest sense. Text can be extracted from a PDF, but the structure, identifiers, references and requirements are not explicit. Machine-readable regulation makes those explicit so software can rely on them.
Who publishes machine-readable regulation today?
Several legislatures and official publishers provide structured formats such as Akoma Ntoso or national XML schemas, and the EU publishes legal data through EUR-Lex and CELLAR. Coverage and depth vary widely, which is why organisations often build their own structured representations.
Why does machine-readable regulation matter for AI?
AI systems that answer from unstructured text cannot show which provision justified an answer. A structured representation gives AI systems and their users stable references, so answers can be checked and reasoning can be traced to its source.
Related topics
- What is machine-executable regulation?
Machine-executable regulation expresses requirements as formal rules software evaluates deterministically, with provenance to the source provision.
- What is LegalRuleML?
LegalRuleML is the OASIS standard for machine-readable legal rules: deontic modalities, defeasibility, temporal validity, source links. How NormLogic uses it.
- What is Regulatory Engineering?
Regulatory Engineering turns regulatory requirements into structured, machine-readable, testable and executable representations. Definition, scope and practice.
- NormLogic, the platform
How Sireto applies Regulatory Engineering to machine-readable and executable regulation.
Which regulation would you make machine-readable first?
We can show how a regulation you work with becomes a structured, traceable knowledge model.