SOURCE
Where did the information come from, and which source is being relied on?
+Definition / Data Authority
DATA AUTHORITY: WHICH DATA CAN AI TRUST, AND WHY?
In the DataAuthority framework, Data Authority is the ability to justify which data can be treated as a reference in a specific context — by keeping its source, time, transformations, coverage, contradictions and limits visible.
01 / The short definition
AUTHORITY IS NOT JUST WHERE DATA IS STORED. IT IS WHY A FACT MAY BE BELIEVED.
The phrase data authority is used in more than one way across industry, governance and public administration. DataAuthority does not claim that its definition is a universal standard. Our framework uses the term for a practical problem: when several sources, transformations or AI systems are involved, how can a person or machine know which statement is justified?
A system can contain an official value and still leave important questions unanswered. Is the value current? Was the source complete? Was a category changed? Did a join drop records? Does zero mean zero, or does it mean “not observed”? Data Authority begins where those questions become visible instead of being silently flattened.
02 / Six questions
A FACT SHOULD TRAVEL WITH THE CONDITIONS THAT MAKE IT TRUSTWORTHY.
TIME
When was it valid, collected or last refreshed?
TRANSFORMATION
What joins, filters, mappings or conversions changed its form?
COVERAGE
What population, territory, period or source set is actually represented?
CONTRADICTION
Do other sources disagree, and is that disagreement preserved?
LIMITS
What is unknown, missing or explicitly unsafe to infer?
03 / Related concepts
DATA AUTHORITY DOES NOT REPLACE GOVERNANCE, LINEAGE OR QUALITY.
Those concepts solve different parts of the problem. In our framework, Data Authority sits across them and asks whether the resulting fact is sufficiently justified for a particular use.
- System of record identifies the designated system for a domain. Data Authority asks why a value from that system is valid in the present context.
- Data governance defines roles, policies, accountability and controls. Data Authority asks whether the governed information can actually support the claim being made.
- Data lineage traces movement and transformations. Data Authority uses that trace to evaluate what changed and what may have been lost.
- Data quality tests properties such as completeness or validity. Data Authority also needs coverage, temporal context, contradictions and explicit limits.
- Ontology / semantic layer organizes meaning and relationships. Data Authority adds the evidential question: what makes a specific fact defensible?
04 / AI
AN AI NEEDS MORE THAN ACCESS TO DATA.
An AI system can retrieve, summarize and combine information. That does not by itself tell the system which source should prevail, whether a dataset is complete enough for the question, or whether a missing value may safely be converted into a zero.
For AI, Data Authority therefore includes machine-readable restraint: preserving unknown states, surfacing conflicts and carrying instructions about what should not be inferred.
05 / Product
THE DATAAUTHORITY ENGINE PUTS THIS INTO PRACTICE.
The public product name is DataAuthority Engine. It currently focuses on structured data — CSV, Parquet, DuckDB and NDJSON — with provenance, checks and an architecture of traceable facts: fact units tied to their source and limits. The purpose is to keep a result attached to the source, transformations, coverage and limits that support it.
The engine is not open for self-service. We currently use it on selected cases and on datasets submitted on request.
We do not present the engine as finished, certified, or infallible. Capabilities are described only at the level of what we can actually show.
06 / Important distinction
DATAAUTHORITY.ORG IS NOT A PUBLIC AUTHORITY OR CERTIFICATION BODY.
“Data authority” can also refer to formal organizational roles or legal authority over data. DataAuthority.org is an independent project and service. The framework described here is our own practical use of the term and should not be confused with a regulator, a Data Protection Authority, or an official standard.