Kate Crawford

Kate Crawford is the Australian-born scholar and artist-researcher working on the political economy and material infrastructure of artificial intelligence — Senior Principal Researcher at Microsoft Research New York, Research Professor at the University of Southern California, and Director of the Knowing Machines Project — the Sloan-Foundation-funded transatlantic research project tracing the histories and politics of how machine-learning systems are trained (see Person entry). She is tracked here as a Voice because her sustained named public output — Atlas of AI (2021) and the art-research collaborations Anatomy of an AI System (2018) and Calculating Empires (2023–2025); the Knowing Machines research programme's ongoing dataset-politics outputs; the February 2024 Nature article on AI's hidden environmental costs; and the keynote circuit running through CHI 2024 and the 2025 Long Now — collectively carries the working frame that AI is a technology of extraction entangled with governance, empire, and power, whose costs and beneficiaries must be named before the ethics conversation can begin.

Voice anchors

The Voice anchors three registers the corpus's voices slice had previously left underweight.

• The extractive / material AI-critique voice anchor entirely. The corpus's algorithmic-accountability voices are anchored on the algorithmic-bias-and-harm register: Timnit Gebru (DAIR, Stochastic Parrots, LLM-critique research), Joy Buolamwini (Algorithmic Justice League, Gender Shades, facial-recognition-bias), Safiya Noble (Algorithms of Oppression, search-engine bias and Black-feminist critique), and Ruha Benjamin (Race After Technology, New Jim Code). None of these occupies the specific register Crawford anchors: the power-and-extraction, material-infrastructure, planetary-costs analysis of AI — the mineral supply chains, the labour discipline of data-annotation platforms and fulfilment warehouses, the energy and emissions of data-centre proliferation, the dataset-construction politics of ImageNet and its successors. Without this Voice entry, the corpus's most load-bearing single publication on that register — Atlas of AI — had no corresponding voice carrying the register forward into the Knowing Machines Project, the Calculating Empires exhibition, and the 2024 Nature article.
• The Australian researcher voice sub-type. The corpus's existing academic voices are concentrated in US geography. Crawford's Australian origin, her PhD from the University of Sydney, and her continuing Honorary Professor relationship there, alongside the École Normale Supérieure Paris visiting chair, anchor a distinct non-US Anglophone research register that the corpus had previously left empty. The Australian critical-humanities tradition — and the vantage it offers from outside both Silicon Valley boosterism and Washington DC policy-circuit framing — is a structurally distinct contribution to the make-AI-good public register.
• The art-research-at-museum-scale voice sub-type. The corpus's voices reach public audiences primarily through academic publications, litigation, and coalition organising. Crawford's art-research practice — most centrally Anatomy of an AI System (2018, MoMA permanent collection; V&A; Ars Electronica Center; Design Museum London; Beazley Design of the Year 2019) and Calculating Empires (2023–2025, Silver Lion Venice Architecture Biennale 2025; European Commission Grand Prize for art and technology; Boghossian Art Prize 2024; shown at Fondazione Prada Milan, Mori Museum Tokyo, Jeu de Paume Paris, Design Museum Barcelona, KW Institute Berlin) — reaches audiences through exhibition venues, museum permanent collections, and international biennial circuits that are structurally distinct from every other corpus voice's public-output channels. The Voice closes the art-research-at-museum-scale voice sub-type entirely.

Signature framings

Four framings recur across Crawford's public output and have done the most work in the make-AI-good movement's public vocabulary.

• AI as extraction: minerals, labour, data, energy — "dehumanizing extractive practices." The central move of Atlas of AI — formalised in Crawford's own framing that AI relies on "dehumanizing extractive practices" and that the technology reflects "power systems benefiting few at the expense of many" — is the most-cited single reframing of the AI-ethics conversation in the corpus. The Yale publisher copy formalises the argument as AI being a technology of extraction "from the minerals drawn from the earth to the labor pulled from low-wage information workers". Karen Hao's 23 April 2021 MIT Technology Review feature — framed as "Stop talking about AI ethics. It's time to talk about power" — carries Crawford's most-propagated single line: AI "is neither artificial nor intelligent — we're just looking at forms of statistical analysis at scale" inheriting the problems of its training data, with the closing call to "contend with the environmental footprint of the systems and the very real forms of labor exploitation". The framing supplies the conceptual backbone on which the corpus's data-centre-organising arcs — the MediaJustice and Kairos campaign and publication entries — rest.
• "Metabolic rifts" — technologies that extract more than they regenerate. Crawford's 2025 Long Now "Mapping Empires" lecture extended the extraction framing into a genealogical argument: tracing an historical arc from Renaissance perspective to AI image models, she introduced the framing that "computation has always been entangled with governance, empire, and expansion" and, drawing on Liebig's critique of industrial agriculture "robbing" soil nutrients and Faraday's latex insulation that devastated rubber forests, installed the concept of the "metabolic rift" — technologies that "extract more than they regenerate" — as the through-line connecting five centuries of extractive technological projects to contemporary AI's energy and emissions infrastructure. The framing places AI's environmental problem inside a long history of colonial resource extraction rather than in the narrower frame of carbon accounting.
• AI's hidden environmental costs. Crawford's February 2024 Nature article "Generative AI's environmental costs are soaring — and mostly secret" named the specific secrecy problem: major AI companies have stopped publishing transparent environmental reporting even as their energy use and water consumption have grown dramatically, and standardised mandatory disclosure is a prerequisite for meaningful governance. The article became a widely-cited anchor for the movement's argument that the true environmental cost of generative AI is being actively concealed from public and policymakers — the publication-side anchor for the same argument the Kairos campaign-and-publication layer carries on the grassroots-organising side.
• The genealogy of technology and power. Calculating Empires: A Genealogy of Technology and Power since 1500 — the 24-metre-long elliptical fresco created with Vladan Joler (opened Fondazione Prada, Milan, November 2023; Silver Lion at Venice Architecture Biennale 2025; European Commission Grand Prize for art and technology; Boghossian Art Prize 2024) — traces colonialism, militarization, and automation from 1500 to 2025 as a single continuous genealogy. The installation argues that the power structures through which today's AI systems operate are not a break from history but a continuation of the same patterns of imperial accumulation, classification, and control that have governed the relationship between technology and power for five centuries. The earlier Anatomy of an AI System (2018, with Vladan Joler) — the supply-chain map of a single Amazon Echo, showing the minerals, labour, data, and logistics the device depends on — was the visual precursor: acquired by MoMA and held in the permanent collections of the V&A, Ars Electronica Center, and Design Museum London.

Public output and venues

Crawford's public-facing work runs through four overlapping channels.

• Books and foundational publications. Atlas of AI (Yale University Press, 2021) is the primary corpus anchor — 336 pages, translated into twelve languages and winner of three international prizes, including the 2022 Sally Hacker Prize from the Society for the History of Technology and the 2022 ASIS&T Best Information Science Book of the Year, and named among the Financial Times best books of 2021. The February 2024 Nature article "Generative AI's environmental costs are soaring — and mostly secret" is the flagship post-2021 publication-side output, anchoring the AI-environmental-transparency argument in the world's most-read scientific journal. The Knowing Machines Project's academic output — including "The Social Construction of Datasets" (with Will Orr, New Media & Society, 2023) and "Building Better Datasets" (with Will Orr, Journal of Data-Centric Machine Learning Research, 2024) — extends the dataset-politics argument into the technical-research literature.
• Art-research collaborations at museum and biennial scale. Anatomy of an AI System (2018, with Vladan Joler) — the supply-chain map of the Amazon Echo that became the visual grammar of the material-AI-critique register — runs in the permanent collections of MoMA, the V&A, the Ars Electronica Center, and the Design Museum London, and received the Beazley Design of the Year 2019. Calculating Empires (2023–2025, with Vladan Joler) won the Silver Lion at the 2025 Venice Architecture Biennale, the European Commission Grand Prize for art and technology, and the Boghossian Art Prize 2024, and has toured to Fondazione Prada Milan, Mori Museum Tokyo, Jeu de Paume Paris, Design Museum Barcelona, Rijksmuseum Twenthe, and KW Institute Berlin — a cumulative exhibition circuit with audiences across six countries.
• International keynote and lecture circuit. Crawford's keynote register spans the academic-and-public spectrum: the Opening Plenary of ACM CHI 2024 in Honolulu — the flagship annual conference in human-computer interaction research; the 2025 Long Now Foundation "Mapping Empires" lecture; and the wider policy-forum circuit through which her Atlas of AI and Knowing Machines work reaches AI-governance audiences. The TIME 100 Most Influential People in AI 2023 recognition anchors the international-influence register at public-mainstream scale.
• Policy influence and framework input. Crawford's work has informed foundational governance frameworks including the EU AI Act, the White House AI Bill of Rights Blueprint, and UNESCO's Recommendations on the Ethics of Artificial Intelligence — giving her voice a direct institutional-policy throughput that most of the corpus's activist-side voices reach only indirectly. This policy-influence register runs through the AI Now Institute's research output, through Crawford's direct advisory relationships with US and European policymakers, and through the Knowing Machines Project's published legal and policy toolkits on training data, dataset deprecation, and generative-AI copyright.

Organisational vehicles

Crawford's public output runs through three principal organisational vehicles. The Knowing Machines Project — the Sloan-Foundation-funded transatlantic research project she directs at USC, "a research project tracing the histories, practices, and politics of how machine learning systems are trained to interpret the world" — is her primary current research vehicle, whose outputs include the Calculating Empires exhibition, Models all the Way Down visual investigation, Synthetic Media and Bird in hand collections, the Generative AI Legal Explainer, A Critical Field Guide for Working with Machine Learning Datasets, the Knowing Legal Machines collection, 9 ways to see a Dataset essays, and the Critical Dataset Studies reading list. The AI Now Institute at NYU — which Crawford co-founded and formerly directed as director of research — is the institutional vehicle through which her policy-influence register and earlier algorithmic-accountability research reached policymakers and the public. Microsoft Research New York (the Social Media Collective, where she is Senior Principal Researcher) provides the institutional home for her peer-reviewed research output.

Why this is a Voice entry

A Voice entry is created here, rather than additional structure on the Person entry, because Crawford's public-facing output is itself the load-bearing object the corpus needs to track: Atlas of AI and the extraction argument it has installed in the AI-governance and grassroots-organising discourse; the Anatomy of an AI System and Calculating Empires art-research collaborations operating at museum and biennial scale; the February 2024 Nature article naming AI's hidden environmental costs; the Knowing Machines Project's ongoing dataset-politics research outputs; the keynote register running through CHI 2024 and the 2025 Long Now; and the policy-influence record running into the EU AI Act, White House AI Bill of Rights Blueprint, and UNESCO's AI ethics recommendations. The corpus's algorithmic-accountability voices had strong coverage on the algorithmic-bias-and-harm register (Gebru, Buolamwini, Noble, Benjamin) but no voice carrying the extractive / material / planetary-costs register — the specific framing that AI is a technology of extraction from minerals, labour, data, and energy, entangled with governance and empire, whose costs are deliberately hidden from the public — until now. Affiliation, training, and biographical detail are recorded on the linked Person entry per the corpus's Person/Voice split.