# AI Now Institute on the hidden material and human work behind AI [Canonical reading](https://escapemeridian.com/readings/ai-now-institute-on-the-hidden-material-and-human-work-behind-ai-42c743f5) AI Now Institute published Yung Au’s guest essay on December 10, 2021, arguing that AI/ML systems are not detached, immaterial products. Au describes how their production and operation involve infrastructure, raw materials, workers and users across different places, even when systems are sold as easy-to-use services. She argues that focusing on where a system is assembled can obscure the labor and expertise involved earlier in its development, including data labeling and cleaning. She also calls attention to who owns data pipelines, which jurisdictions govern systems, and who can hold providers accountable. In her account, the power to frame and benefit from AI/ML production is less widely shared than the intelligence and labor involved in it. Au argues that labels for AI/ML systems should help make their supply chains, environmental effects, labor conditions and accountabilities more visible. The essay raises questions about how to describe those connections without obscuring them, and how to make systems answerable to the people affected. It does not offer a tested labeling method or assess the effects of a particular policy. Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims. ## Original source AI Now Institute · Article Original source date: Dec 10, 2021 (2021-12-10T15:49:38Z). Date basis: Date from the publisher’s own records, through its API. Added to Escape Meridian: 2026-10-29T16:18:32.667Z. First observed: 2026-10-20T23:00:48.791Z. Retrieved: 2026-10-28T09:09:52.603Z. - [A New AI Lexicon: Exporting AI](https://ainowinstitute.org/publications/collection/a-new-ai-lexicon-exporting-ai) ## Limits Text as retrieved on October 28, 2026; originally published December 10, 2021. The page is marked as edited after publication; the retrieved text may include later additions, which are not treated here as part of the original essay. The essay raises questions about labeling and accountability but does not evaluate a specific labeling method or policy. ## Open question What kinds of labels or accountability arrangements could make AI/ML systems’ labor, environmental effects and jurisdictions visible without obscuring them? ## Source-stated Yung Au argues that AI/ML systems are often presented as immaterial despite their physical and human dependencies. > Software and virtual products are often decoupled from their material entanglements - [Supporting source](https://ainowinstitute.org/publications/collection/a-new-ai-lexicon-exporting-ai) ## Source-stated Au says that AI/ML work and expertise extend beyond the places where systems are presented as assembled. > Labor, decisions, and expertise are needed at each stage prior to a model’s deployment, and even still after it is deployed. - [Supporting source](https://ainowinstitute.org/publications/collection/a-new-ai-lexicon-exporting-ai) ## Source-stated Au argues that power over AI/ML production is less widely distributed than the intelligence involved in the process. > While intelligence might be everywhere in the process, the power to frame, gate-keep and benefit off of production of AI/ML systems is less distributed. - [Supporting source](https://ainowinstitute.org/publications/collection/a-new-ai-lexicon-exporting-ai) ## Source-stated Au calls for labels and discussion that make AI/ML systems’ material and geographic connections more visible. > It becomes increasingly important then, to converse about ways to better embed geographies in discussions on AI/ML and its governance. - [Supporting source](https://ainowinstitute.org/publications/collection/a-new-ai-lexicon-exporting-ai) ## Source and contributor context AI Now Institute · Publisher The publisher is identified here. No individual author has been inferred. ## Source media Media retrieval status: unchecked. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## Machine learning Methods that let computers learn patterns from examples instead of following rules written by hand. - [Definition reference](https://developers.google.com/machine-learning/glossary#machine-learning) ## AI Now Institute AI Now Institute studies the power and institutions shaping AI and advocates for public accountability. - [Explore this context](/entities/ai-now) - [Reference](https://ainowinstitute.org/) ## Editorial provenance Version 1. Publication ID: 69fc5ef2-9a20-0990-833b-df34450d91c8. Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5. Source check: passed by anthropic/claude-opus-5.5 on 2026-10-29T16:18:31.925Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: bridge\_admitted\_by\_model, second\_opinion\_categorical. Observation: a5c75f760c49d115a05ef4e0b44dbaaf916f504ea4eed752b8d107fa740fbb2b. Source text hash: 867a28e0bdfb720ec34b2eb4627c7dc7c3feaf2b38a57a1f6bb1b736cf61b315. ## Revision and source-check history ### Version 1 — 2026-10-29T16:18:32.667Z AI Now Institute published Yung Au’s guest essay on December 10, 2021, arguing that AI/ML systems are not detached, immaterial products. Au describes how their production and operation involve infrastructure, raw materials, workers and users across different places, even when systems are sold as easy-to-use services. She argues that focusing on where a system is assembled can obscure the labor and expertise involved earlier in its development, including data labeling and cleaning. She also calls attention to who owns data pipelines, which jurisdictions govern systems, and who can hold providers accountable. In her account, the power to frame and benefit from AI/ML production is less widely shared than the intelligence and labor involved in it. Au argues that labels for AI/ML systems should help make their supply chains, environmental effects, labor conditions and accountabilities more visible. The essay raises questions about how to describe those connections without obscuring them, and how to make systems answerable to the people affected. It does not offer a tested labeling method or assess the effects of a particular policy. ## Original source AI Now Institute · Article Original source date: Dec 10, 2021 (2021-12-10T15:49:38Z). Date basis: Date from the publisher’s own records, through its API. Added to Escape Meridian: 2026-10-29T16:18:32.667Z. First observed: 2026-10-20T23:00:48.791Z. Retrieved: 2026-10-28T09:09:52.603Z. - [A New AI Lexicon: Exporting AI](https://ainowinstitute.org/publications/collection/a-new-ai-lexicon-exporting-ai) ## Limits Text as retrieved on October 28, 2026; originally published December 10, 2021. The page is marked as edited after publication; the retrieved text may include later additions, which are not treated here as part of the original essay. The essay raises questions about labeling and accountability but does not evaluate a specific labeling method or policy. ## Open question What kinds of labels or accountability arrangements could make AI/ML systems’ labor, environmental effects and jurisdictions visible without obscuring them? ## Source-stated Yung Au argues that AI/ML systems are often presented as immaterial despite their physical and human dependencies. > Software and virtual products are often decoupled from their material entanglements - [Supporting source](https://ainowinstitute.org/publications/collection/a-new-ai-lexicon-exporting-ai) ## Source-stated Au says that AI/ML work and expertise extend beyond the places where systems are presented as assembled. > Labor, decisions, and expertise are needed at each stage prior to a model’s deployment, and even still after it is deployed. - [Supporting source](https://ainowinstitute.org/publications/collection/a-new-ai-lexicon-exporting-ai) ## Source-stated Au argues that power over AI/ML production is less widely distributed than the intelligence involved in the process. > While intelligence might be everywhere in the process, the power to frame, gate-keep and benefit off of production of AI/ML systems is less distributed. - [Supporting source](https://ainowinstitute.org/publications/collection/a-new-ai-lexicon-exporting-ai) ## Source-stated Au calls for labels and discussion that make AI/ML systems’ material and geographic connections more visible. > It becomes increasingly important then, to converse about ways to better embed geographies in discussions on AI/ML and its governance. - [Supporting source](https://ainowinstitute.org/publications/collection/a-new-ai-lexicon-exporting-ai) ## Source and contributor context AI Now Institute · Publisher The publisher is identified here. No individual author has been inferred. ## Source media Media retrieval status: unchecked. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## Machine learning Methods that let computers learn patterns from examples instead of following rules written by hand. - [Definition reference](https://developers.google.com/machine-learning/glossary#machine-learning) ## AI Now Institute AI Now Institute studies the power and institutions shaping AI and advocates for public accountability. - [Explore this context](/entities/ai-now) - [Reference](https://ainowinstitute.org/) ## Editorial provenance Version 1. Publication ID: 69fc5ef2-9a20-0990-833b-df34450d91c8. Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5. Source check: passed by anthropic/claude-opus-5.5 on 2026-10-29T16:18:31.925Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: bridge\_admitted\_by\_model, second\_opinion\_categorical. Observation: a5c75f760c49d115a05ef4e0b44dbaaf916f504ea4eed752b8d107fa740fbb2b. Source text hash: 867a28e0bdfb720ec34b2eb4627c7dc7c3feaf2b38a57a1f6bb1b736cf61b315. ### passed — 2026-10-29T16:18:31.925Z — anthropic/claude-opus-5.5 Publication: 69fc5ef2-9a20-0990-833b-df34450d91c8 attribution — pass: The reading correctly attributes the essay to Yung Au as a guest post published by the AI Now Institute as part of its AI Lexicon project. The title's framing of AI Now as publisher is consistent with the summary's attribution of the arguments to Au. > A guest post by Yung Au. Yung is a doctoral researcher at the Oxford Internet Institute (OII) summary — pass: The summary is readable without specialist background. It names the actor and the core argument: AI is not immaterial, labor is distributed, power is concentrated, and supply chains should be made visible. It fairly states that the essay offers no tested labeling method. AI/ML and data pipelines are not spelled out, but this does not block understanding. The question follows from the essay's own closing questions. > How do we assess whether an AI/ML system or its end products uses fair labor, to which jurisdiction it is accountable to, or what their environmental effects are? date — pass: The original publication instant of 2021-12-10T15:49:38Z gives the UTC day December 10, 2021, which matches the reading at day precision. The page was edited after publication (2025 dateModified), and the reading discloses this. The cited passages are core essay body and plainly part of the original piece. The reading uses no hindsight framing and presents no later events. The evidence kind is a publisher page, so no archive disclosure is needed. support — pass: Each claim is directly supported by verbatim essay text: the immateriality framing, labor at every stage, power being less distributed than intelligence, and the call to embed geographies and better labeling. The summary's points about accountability and jurisdiction are source-stated. It adds no unsupported benefits or outcomes. > While intelligence might be everywhere in the process, the power to frame, gate-keep and benefit off of production of AI/ML systems is less distributed. media — pass: No media was selected and no verified media candidates were supplied, so the null media choice is appropriate. ## Recorded references Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter. No public references recorded here yet. [Reference data (JSON)](/readings/42c743f5855943d19d07a5124b384cc11a03b4f8894ddf72f910c40cc3896c6d/references.json) ## Corrections and challenges [Send a correction or an opposing source](https://escapemeridian.com/corrections/) [JSON record](https://escapemeridian.com/readings/ai-now-institute-on-the-hidden-material-and-human-work-behind-ai-42c743f5.json)