Labor Vector
Labor Vector meta-aggregator / Labor producing AI

Responsible Sourcing Across the Data Supply Line

Labor producing AIGains and lossesworkstream and resource libraryactive
Responsible Sourcing Across the Data Supply Line: a workstream and resource library: recommendations and practitioner tools (guidelines, checklist, templates, a case study, a white paper). Operator: Partnership on AI (PAI), AI, Labor, and the Economy programme; the guidelines developed in partnership with DeepMind. Its unit: a guideline or practical tool addressed to AI practitioners (five guidelines; six resources in the library). Whose labor: data enrichment work: data preparation and cleaning (annotation, intent recognition, sentiment analysis, image recognition, speech-to-text validation) and human-review work (content moderation, feedback loops, validating algorithmic outputs and models).
Visit Responsible Sourcing Across the Data Supply Line ↗The /non rowThe DSL audit
the Responsible Sourcing workstream aims to develop recommendations and actionable resources to improve the working conditions of these professionals.

In its own words: https://partnershiponai.org/workstream/responsible-sourcing/, read 2026-10-29

Who supplies, who directs, who gains

Supplies the labordata enrichment workers, described as faraway workers under precarious conditions
Directs itAI practitioners (data scientists, AI engineers, product managers) who design data enrichment projects and set payment terms, and the vendors and platforms they source through
Receives the gains, bears the costsnot recorded as a quantity; the sources state that the devaluing of this work brings poor working conditions for workers and worse outcomes for AI development

Each as the resource itself records it. Where it records nothing, the census says so.

What it records

IntersectionsLabor producing AI: annotation, moderation, evaluation, maintenance, and creative or scholarly work entering training and retrieval; Distribution of gains and losses: wages, employment, workload, credit, ownership, and who captures productivity gains
Kind of evidencepractitioner guidance
Unit of recorda guideline or practical tool addressed to AI practitioners (five guidelines; six resources in the library)
Whose labordata enrichment work: data preparation and cleaning (annotation, intent recognition, sentiment analysis, image recognition, speech-to-text validation) and human-review work (content moderation, feedback loops, validating algorithmic outputs and models) op
Methodguidance drafted from PAI's white paper, developed with DeepMind and tested and adopted by DeepMind organization-wide; feedback invited from practitioners, workers and the AI community
Coveragegeography: data enrichment workers around the world (not otherwise specified); period: workstream posts from December 2020 to April 2026; library and guidelines published November 2022
What it findsthe sources hold recommendations and report no measured findings of their own; the library states that research shows data enrichment workers often face inconsistent and inappropriate compensation, unclear instructions, lack of recognition and stress from long, ad-hoc hours; guideline 1: pay all data enrichment workers above the local living wage, including time spent reading instructions and in training
Its stated limitsthe guidelines were tested at one AI developer (DeepMind); the library asks for feedback on how they can be improved; PAI cautions against setting per-task rates from a simple median completion time

op the resource's own statement · 3p a third-party account · obs our observation · rd our reading. Hover a tag for its quote; each opens below with its source.

What this evidence can settle

Can answer

  • what PAI recommends AI practitioners do when sourcing enriched data (pay, pilots, worker selection, instructions, communication)
  • what tools PAI provides to compare vendors and look up local living wages

Cannot answer

  • what data enrichment workers are actually paid
  • which AI developers follow the guidelines, beyond DeepMind
  • whether the guidelines changed conditions for workers (the DeepMind case study was not read)

In the worked questions

Where it meets Labor Vector

rd none stated. The archive reads guideline 1's instruction to count time spent reading instructions, training and reviewing work before submission as the resource's own accounting of uncompensated labor around a task; it stays in PAI's terms.

Entity

Canonical nameResponsible Sourcing Across the Data Supply Line
Also known asResponsible Sourcing workstream, Responsible Data Enrichment Sourcing Library, Improving Conditions for Data Enrichment Workers, Data Enrichment Sourcing Guidelines
Addresshttps://partnershiponai.org/workstream/responsible-sourcing/
OperatorPartnership on AI (PAI), AI, Labor, and the Economy programme; the guidelines developed in partnership with DeepMind
Founded2020 (the earliest workstream post, 22 December 2020)
Kindworkstream and resource library: recommendations and practitioner tools (guidelines, checklist, templates, a case study, a white paper)
Lifecycleactive — workstream updates from 22 December 2020 to 29 April 2026; the library and DeepMind collaboration announced 16 November 2022; vendor engagement guidance and transparency template introduced August 2024, put to public comment on 5 September 2024, and released in finalized versions in PAI's post of 29 April 2026; OECD.AI's catalogue (third party) classifies the library's maturity as implemented in multiple projects
Countfive guidelines; six resources listed in the library (as of 2026-10-29) op
Data accessguidelines, checklist and white paper as PDFs; vendor comparison and living-wage templates as Google Sheets, linked from the library page
Licencenot stated in PAI's saved pages (site footer: © 2026 Partnership on AI | All Rights Reserved); OECD.AI's catalogue entry for the library, a third party, classifies its usage rights as Open source/Permissive op
Versions and datesfirst post: 2020-12-22; library: 2022-11-16; vendor guidance: 2024-08-28; public comment: 2024-09-05; vendor guidance finalized: 2026-04-29; latest post: 2026-04-29; read: 2026-10-29
Observation2026-10-29: read

Relations

Captures: the composition layer's account

The /non row

Evidenceevery field's quote, its source and the hash of the saved copy25
Negative tagswhat it is not3
  • NOT a dataset or survey of data enrichment workers
  • NOT a rating or certification of vendors
  • NOT an audit of DeepMind's adoption: the case study is DeepMind's account with PAI
Sourcesevery source read, with its date and hash5
Not yet checkedwhat remains to read5
  • the white paper Responsible Sourcing of Data Enrichment Services and its Annex 1
  • the DeepMind case study
  • the PDFs and Google Sheets templates
  • the workstream's blog posts other than the post of 29 April 2026
  • the vendor engagement guidance and transparency template themselves (their finalized versions are announced in the post of 29 April 2026; the documents were not opened)

Cite this record

Lee Sharks, ed. "Responsible Sourcing Across the Data Supply Line." Labor Vector meta-aggregator, laborvector.org, compiled 2026-10-29. https://www.laborvector.org/aggregators/pai-responsible-sourcing

Record https://www.laborvector.org/aggregators/pai-responsible-sourcing · entity https://www.laborvector.org/aggregators/pai-responsible-sourcing#entity · the census as data

@misc{lv-pairesponsiblesourcing2026,
  editor = {Sharks, Lee},
  title = {Responsible Sourcing Across the Data Supply Line},
  howpublished = {Labor Vector meta-aggregator, laborvector.org},
  year = {2026},
  note = {Compiled 2026-10-29; every field sourced and dated},
  url = {https://www.laborvector.org/aggregators/pai-responsible-sourcing}
}
Editor: Lee Sharks (ORCID 0009-0000-1599-0703). This page is the Labor Vector meta-aggregator's record of the resource above: its reading and fields are ours, and each field carries its source, its basis and the hash of the copy read. The resource's own terms and findings are kept as its own. Compiled 2026-10-29. CC BY 4.0.