Provenance and Governance of AI Supply Chains

AI systems benefit if we have a usable account of where data came from. Provenance, licensing, audits, counterfactuals, and attestation are all potential ways to improve the supply chain.

Key References

  1. 2026 · ProjectData Counterfactuals
  2. 2025 · ProjectDataLicenses.org

More work in this topic

A not-totally-exhaustive (because there’s a lot of overlap between these themes) list of “outputs” that meet this theme.

2026May 2026ShortpostAugmentation is a data flow problemData Leverage Short Posts
2026Apr 2026BlogAttestation across the AI Supply ChainData Leverage
2026Mar 2026BlogTwo natural allies of a "Data Transparency" agenda: capabilities forecasters and social simulatorsData Leverage
20262026ProjectData Lava Lamp
2025Nov 2025BlogAlmost Everybody -- Including Both Data Creators and AI Companies -- Stands to Benefit from Clearer "Data Rules".Data Leverage
2025Oct 2025PaperResponsible AI in the OSS: Reconciling Innovation with Risk Assessment and DisclosureAIES
2025Sep 2025BlogWhich datasets should we assume are "in all the AI models"?Data Leverage
2025Apr 2025MediaInside Meta's secret experiments that improve its AI modelsBusiness Insider
2025Jan 2025BlogAI Labs Should Open Source Data Protection TechnologiesData Leverage
2025Jan 2025MediaOpenAI is reaping what it sowed with DeepSeek. What's that old saying about karma?Business Insider
2025Jan 2025BlogLive by the free-content-for-training sword, die by the free-content-for-training swordData Leverage
2025Jan 2025WorkshopResponsible AI in Open Ecosystems: Reconciling Innovation with Risk Assessment and DisclosureAAAI 2025 Workshop on AIGOV
2024Dec 2024BlogSelling AGI like AG1: Will Consumers Push Back Against Proprietary Blends of Herbs and of Data?Data Leverage
2024Sep 2024TalkHuman-Centered AI Research and New Paradigms for Generative AI DataHuman-Centered AI Conference @ Pepperdine University
2024Aug 2024Other paperA step forward in tracing and documenting dataset provenanceNature Machine Intelligence, News & Views
2024Jun 2024MediaIs data supply AI’s Achilles’ heel?UBS
2024Mar 2024MediaJews have always been prolific writers. Has AI wound up with too much of their work?Jewish Telegraphic Agency
2024Mar 2024MediaNicholas Vincent explains why robots.txt is no longer enough to protect against web scrapingIT Brew
2023Nov 2023BlogBuilding a Data Pipeworks for Democratic AI: From Human Knowledge to Records to AI SystemsData Leverage
2023Oct 2023WorkshopCan Licensing Mitigate the Negative Implications of Commercial Web Scraping?Workshop at CSCW 2023
2023Mar 2023TalkCommunity Dialogue on Accountable Governance and DataCommunity Data Science Collective Community Dialogues
2023Mar 2023BlogPlural AI Data AlignmentData Leverage
2023Feb 2023TalkHuman-centered data and language models -- Privacy, data as labor, and licensingStanford Social NLP Reading Group Talks
2022Dec 2022PanelDoes the rise of AI need us to adopt new data licensing policies?Open Health Data and AI Summit 2022
2022Jun 2022PaperBehavioral Use Licensing for Responsible AIACM FAccT
2021Nov 2021Other paperPreview of "Data and its (dis)contents: A survey of dataset development and use in machine learning research"Patterns
2021May 2021PaperAddressing Documentation Debt in Machine Learning Research: A Retrospective Datasheet for BookCorpusNeurIPS Datasets and Benchmarks