Data Provenance Explorer

databaseprovenancemonitoringactive

The Data Provenance Explorer, from the Data Provenance Initiative (Longpre, Mahari and others), traces the sources and licences of AI training datasets: nearly 4,000 text, speech and video datasets by its own count. Its founding audit covered more than 1,800 text datasets and found licence omission above 70% and error rates above 50%.

Visit Data Provenance Explorer ↗The /non row Capture Follow

Entity

Canonical name	Data Provenance Explorer
Address	https://www.dataprovenance.org/
Operator	Data Provenance Initiative (Longpre, Mahari, Chen, Obeng-Marnu and others) op
Kind	database
Roles	provenance, monitoring
Lifecycle	active
Observation	2026-10-28: read
In its own words	“Public measurement infrastructure for AI data, model ecosystems, and real-world AI use” op
Unit	dataset (text, speech or video), traced to its creators, sources, licences, geographies, modalities and derivation chain op
Count	nearly 4,000 text, video and speech datasets (the Dataset Explorer); the founding audit covered more than 1,800 text datasets (as of 2026-10-29) op
Note	the founding audit found 'license omission of 70%+ and error rates of 50%+' on dataset hosting sites. Its second audit, Consent in Crisis (NeurIPS 2024, by the Initiative's home page), read 14,000 web domains and found that 'a full 45% of C4 is now restricted' by terms of service op

op operator statement · 3p third-party report · obs our observation · rd our reading

What it is

The Initiative calls its work "public measurement infrastructure for AI data, model ecosystems, and real-world AI use." The Explorer lets practitioners trace and filter the provenance of popular open fine-tuning collections, back to their sources, creators and licences.

What its audits found

The founding audit (arXiv:2310.16787) found "license omission of 70%+ and error rates of 50%+" on dataset hosting sites. Consent in Crisis (arXiv:2407.14933, NeurIPS 2024) read 14,000 web domains and found that "a full 45% of C4 is now restricted" by terms of service.

Place in the field

It records provenance at the training layer, where the census's reception aggregators record it at the composition layer. The two ends of one chain: what a model was given, and what it credits when it answers.

For a research question

Records attribution?

yes: sources, creators and licences per dataset op

Relations

Captures: the composition layer's account

«data provenance explorer» — Google AI Overview, 2026-10-29
the composition layer's account of the entity, seated in the Capture Registry · observed 2026-10-29

The /non row

its /non row (EA-NEGONT-02 v0.9, equal case)
composed 2026-10-29 against its own pages, its Wikipedia article or its absence, the capture's cards and the archive

Provenanceevery source read, with its date4

https://arxiv.org/abs/2310.16787
read 2026-10-28 · audit
https://www.dataprovenance.org/
read 2026-10-28 · self-description
https://arxiv.org/abs/2407.14933
read 2026-10-28 · Consent in Crisis audit
https://www.alexanarch.org/captures/data-provenance-explorer-aio-20261029/
read 2026-10-29 · the Initiative's own page text, on the Overview's cards (the pages render nothing to this record's fetcher)

Not yet checkedwhat remains to read1

other collections

Cite this record

Lee Sharks, ed. "Data Provenance Explorer." MMRS meta-aggregator, Machine-Mediated Reception Studies, compiled 2026-10-29. https://www.machinemediation.org/aggregators/data-provenance-explorer/

Record https://www.machinemediation.org/aggregators/data-provenance-explorer/ · entity https://www.machinemediation.org/aggregators/data-provenance-explorer/#entity

Editor: Lee Sharks (ORCID 0009-0000-1599-0703). This page is the MMRS record of the entity above: its reading, fields and relations are MMRS's, and each field carries its source and basis. Record https://www.machinemediation.org/aggregators/data-provenance-explorer/ · entity https://www.machinemediation.org/aggregators/data-provenance-explorer/#entity · compiled 2026-10-29