SPXI instrument · method 0.9 of 87, level 0 (Measurement): Capture Registry (longitudinal observation). An instrument of SPXI, the Semantic Packet for eXchange & Indexing; the whole practice is itemized at spxi.dev/methods.

Capture Registry › capture data-provenance-initiative-aio-20261029

One record of the canonical Capture Registry (EA-WG-CAPTURES-01), cited at https://www.alexanarch.org/captures/data-provenance-initiative-aio-20261029/. the canonical Capture Registry (version 12.379) · the address page · this card in the gallery · this record as data · table of contents.

AXN:0ACD.CAPTURE — this address. Each observation below is one AXN, sealed with its analysis and measures; the identifier names the reading together with the response it reads.
Machine Reception2026-10-29
Data Provenance Initiative
Originator: the Data Provenance Initiative (volunteer effort; paper by Shayne Longpre and 16 other authors) · outside the archive (out-universe) · dataset · SPXI treatment: partial
BROAD MATCH
no
image
THE DATA PROVENANCE INITIATIVE, BY NAME, ON GOOGLE'S AI OVERVIEW, SIGNED OUT, 2026-10-29: 'a multi-institutional research project—led in part by researchers from MIT—that audits the datasets, licenses, and web domains used to train artificial intelligence models.' Four platforms (the Data Provenance Explorer over 1,800+ datasets, a Web Consent Explorer, Open Model Evolution, an AI Observatory) and three findings (license omission over 70% and error over 50% on hosting platforms, closed datasets concentrated in creative tasks and lower-resource languages, a rise of non-commercial licences driven by synthetic data). Four cards: the initiative's home page ('4,000+ audited datasets; 14,000 tracked web domains'), the arXiv paper, MIT Sloan, Nature Machine Intelligence (1,858 datasets across 44 collections).
Full record — 4,969 characters, 4 sources
Capture record
captured
2026-10-29
surface
Google AI Overview
auth state
signed out, incognito
evidence class
paste
PER (historical scalar)
0.5
PER units retained
inst, src
citations read
4
observation id
OBS-118fbafa73c9
address id
ADDR-c0bda0b5919d
Reading

The innocent reading first. The Overview identifies the project, lists its tools and reports what its audits found, from its own site, its paper and the press.

Findings
Analysis analyst prose, not machine text

As with the revived Papers with Code on the MMRS side, the composition is ahead of the record on the entity's current extent: the census formalized the initiative from its first release and its repository, and the Overview read its present home page. The capture is a correction to be made from the entity's own page.

Machine text, verbatim
CAPTURE-TIME VERBATIM RECORD — paste of the Overview expanded, with its source cards · complete as supplied · READ IN FULL 2026-10-29
[Google AI Overview at the query, signed out and incognito (operator defaults of 2026-10-01 and 2026-10-14), 2026-10-29. The paste is the Overview expanded (the 'AI Mode Conversation' view, with the source cards that follow the answer); the 'You said' line carries the query twice, run together, and is given once here. Inline links are Google redirect URLs, kept as pasted.] [QUERENT] Data Provenance Initiative [SYSTEM] The [Data Provenance Initiative](https://www.google.com/goto?url=CAESVwHrOzAVIs_bBUxr7ducYJKxi6mO2H8PFkOqeMYxYB9BZOGggLJumXN31RQqX2ghNJ2wlF1SU0MHuuvc_7Cgibm_RylNPmv9oYEqoRYseVd3oqu3XZKm1g) is a multi-institutional research project—led in part by researchers from MIT—that audits the datasets, licenses, and web domains used to train artificial intelligence models. [[1], [2]] Key Tools & Platforms * Data Provenance Explorer: An interactive tool that audits and filters over 1,800+ text and fine-tuning datasets across dozens of collections to inspect sources, lineages, and license types. * Web Consent Explorer: A system that tracks thousands of web domains for consent signals and web policies. * Open Model Evolution: A monitor tracking open-source and open-weight AI models for transparency and concentration. * AI Observatory: A comparative tool for analyzing real-world AI conversation sets. [[1], [2], [3]] Major Findings from Audits * License Omission & Errors: Widely used hosting platforms exhibit over a 70% license omission rate and more than 50% error rates in how they categorize dataset licenses. [1] * Composition Divides: Commercially closed or restrictive datasets often monopolize creative tasks, specialized topics, and lower-resource languages. [1] * Non-Commercial Spikes: There has been a significant rise in non-commercial and academic-only licenses, frequently driven by synthetic data generated via commercial APIs. [1] Would you like to explore how to use the Data Provenance Explorer, or look into specific dataset licenses? Data Provenance Initiative Data Provenance Initiative
Sources (4) as cited, and as the copy produced them
  1. Data Provenance Initiative third_party
    Data Provenance Initiative
    Platforms: Data Provenance Initiative (maps AI datasets, web policies, model markets, real-world use). Metrics: 4,000+ audited datasets; 14,000 tracked web doma...
    as pastedData Provenance Initiative (maps AI datasets, web policies, model markets, real-world use). Metrics: 4,000+ audited datasets; 14,000 tracked web doma... arXiv
  2. arXiv third_party
    [2310.16787] The Data Provenance Initiative: A Large Scale Audit of Dataset Licensing & Attribution in AI
    Audit: 1800+ text datasets traced and audited by legal and machine learning experts to evaluate licensing, lineage, and attribution in AI training. Findings: Co...
    as pasted[2310.16787] The Data Provenance Initiative: A Large Scale Audit of Dataset Licensing & Attribution in AI Audit: 1800+ text datasets traced and audited by legal and machine learning experts to evaluate licensing, lineage, and attribution in AI training. Findings: Co... MIT Sloan
  3. MIT Sloan third_party
    Bringing transparency to the data used to train artificial intelligence | MIT Sloan
    Problem: Untraced AI training datasets cause legal, copyright, regulatory, bias, and quality risks due to poor documentation and high license error rates (>50%)
    as pastedBringing transparency to the data used to train artificial intelligence | MIT Sloan Problem: Untraced AI training datasets cause legal, copyright, regulatory, bias, and quality risks due to poor documentation and high license error rates (>50%) Nature
  4. Nature third_party
    A large-scale audit of dataset licensing and attribution in AI | Nature Machine Intelligence
    Audit: 1,858 text/finetuning datasets across 44 collections evaluated for licensing, attribution, and lineage via the Data Provenance Initiative and DPExplorer.
    as pastedA large-scale audit of dataset licensing and attribution in AI | Nature Machine Intelligence Audit: 1,858 text/finetuning datasets across 44 collections evaluated for licensing, attribution, and lineage via the Data Provenance Initiative and DPExplorer.
  5. unattributed segment no cited source matches this text
    Platforms:
↻ Re-runpermalink