Skip to main content

Data provenance

Documentation of where a piece of data originally came from, who owned it, and who has handled it, proving it is genuine and untampered.

The Simple Version

Documentation of where a piece of data originally came from, who owned it, and who has handled it, proving it is genuine and untampered.

Detailed Explanation

Data provenance answers the question 'where did this data come from and can we trust it?' It differs from lineage in scope: lineage describes the full transformation journey, while provenance focuses on origin and custody. In AI contexts, provenance is important for establishing that training data was collected lawfully (copyright, GDPR), that it is authentic (not manipulated), and that the organisation has the right to use it. W3C PROV-O provides a vocabulary for representing provenance. The EU AI Act requires providers to document the 'origin' of training data used in high-risk systems.

Key Characteristics

  • Focuses on origin, ownership, and custody, complementary to but distinct from lineage
  • Establishes legal basis for data use (licence, consent, legitimate interest)
  • Supports authenticity verification and manipulation detection
  • Formalised in W3C PROV-O ontology

Why It Matters

Legal and data engineering teams must jointly establish provenance records for AI training data, particularly for data sourced from third-party suppliers, web scraping, or data marketplaces.

Real-World Analogy

Like the provenance certificate accompanying a painting at auction, documenting the chain of ownership from the artist to the current seller to verify authenticity and rule out forgery.

Common Misconceptions

  • Provenance and lineage describe the same thing, provenance is about origin and custody; lineage is about transformation history.
  • Provenance is only relevant for sensitive personal data, it matters for any data where authenticity, licensing, or manipulation could affect AI system reliability.

Related Terms

Related Articles

Sources & Further Reading