◆ Semantic Web Catalogue

Problem thread

Who owns what the crowd built?

When volunteers build a database together, what stops whoever hosts it from closing it, selling it or shutting it down?

issue
governance
also sustainability, adoption & incentives
status
open
steps
8, 1998–2025
threads
all problem threads

Many of the largest databases about films, music and the world were typed in by fans and volunteers. IMDb's fan lists became Amazon's property, and CDDB's contributors saw their work turned into the proprietary Gracenote, which provoked MusicBrainz and its public-domain data. Freebase used an open licence but closed when its owner withdrew, and Wikidata now holds data waived into the public domain and run by a foundation. The same question is back with AI: who benefits from what volunteers built.

Lineage

  1. Apr 1998
    precedent project ★ seminal
    IMDb

    Fan lists from rec.arts.movies become a company in 1996 and are bought by Amazon in 1998. Contributors keep copyright in what they submit, but the compilation becomes IMDb's exclusive property, and the data is released only for non-commercial use. [src1][src2]

  2. 2000
    poses project
    CDDB

    Users had filled CDDB believing it would stay free. Sold in 1998 and renamed Gracenote in 2000, it moved to a proprietary licence; critics said it exploited unpaid contributors, and in March 2001 it banned unlicensed applications. [src]

  3. 17 Jul 2000
    responds project ★ seminal
    MusicBrainz
    ↳ responds to CDDB

    Founded in response to CDDB's restrictions. Its core data has been in the public domain under CC0 since 2003, so no later owner can close it; the MetaBrainz Foundation pays for upkeep by licensing the live data feed. [src]

  4. Mar 2007
    proposes project ★ seminal
    Freebase

    A company-run knowledge base built partly from users' contributions and partly from harvested sources such as Wikipedia and MusicBrainz, its data licensed CC BY so that copies could outlive the host. [src]

  5. 14 Nov 2009
    proposes project
    Archive of Our Own

    Fans build their own archive, run by a fan-run non-profit, the Organization for Transformative Works, rather than depending on a commercial host. [src1][src2]

  6. 29 Oct 2012
    proposes project ★ seminal
    Wikidata

    A volunteer-edited knowledge base hosted by the non-profit Wikimedia Foundation, with all structured data waived into the public domain under CC0. [src]

  7. 11 Apr 2016
    measures writing ★ seminal
    From Freebase to Wikidata: The Great Migration

    When Google closed Freebase, the open licence let its data be offered to Wikidata, but not all of it could be carried over, and the community had to choose what to accept: an open licence kept the data copyable, not alive. [src]

  8. 25 Jun 2025
    poses concept
    CC signals

    AI training raises the question again for commons: Creative Commons proposes signals for dataset holders to state how machines may reuse what they hold. [src]

Recent signals

Dated events that show the problem moving, newest last.

  1. 1 Jul 2025

    Cloudflare starts blocking AI crawlers by default on new sites and offers publishers a way to charge them per crawl.[src]

  2. 17 Oct 2025

    The Wikimedia Foundation reports human pageviews down about 8% on a year earlier, and links the fall to AI answers in search. Wikidata[src]

  3. 15 Jan 2026

    Amazon, Meta, Microsoft, Mistral AI and Perplexity become paying partners of Wikimedia Enterprise, joining Google.[src]

Open now

Under each question, the answers people have put forward or built. The catalogue gives no odds on them.

  • Can a volunteer-built database pay its way without enclosing its data? [src1][src2]
    MetaBrainz keeps MusicBrainz's core data in the public domain and charges commercial users for the live feed; IMDb and Gracenote instead made the data itself the product.
    • The data stays free and heavy users pay for the service. MusicBrainz
      MetaBrainz charges commercial users for MusicBrainz's live feed; Wikimedia Enterprise sells high-volume access to Wikipedia's freely licensed content.[src1][src2]
    • The data itself becomes the product. IMDb
      IMDb and Gracenote, which grew out of CDDB, turned volunteer contributions into licensed commercial databases.[src1][src2]
  • Who should benefit when AI systems are trained on what volunteers built? [src]
    CC signals let holders of datasets state expectations for machine reuse, partly legally binding and partly a social norm.
    • AI companies pay the hosts for access.
      Wikimedia's 2026 Enterprise partners pay for faster, larger access than scraping gives.[src]
    • Holders state machine-readable terms for AI reuse. CC signals
      CC signals let a dataset's holders say how they expect AI systems to give back.[src]
    • Hosts block AI crawlers unless they pay.
      Cloudflare's default block and pay-per-crawl, from July 2025.[src]

Forces

Pressures that push this problem one way or another, as their sources argue them. All forces

How to contribute