Kate Crawford

Early research tracing the material supply chains behind AI: rare earth minerals mined under dangerous conditions, server farms consuming vast energy, data center workers in precarious labor conditions. Challenged the 'cloud' metaphor that obscures AI's physical footprint.

'Halt the Use of Facial-Recognition Technology Until It Is Regulated' (2019) and related work argued that AI systems are not objective mirrors of reality but encode the political and economic interests of their creators. 'Bias' is not a bug to be fixed with more data; it reflects the power structures embedded in the training process.

In 'Atlas of AI' (2021), argued that every AI system rests on three forms of extraction: natural resources (minerals for hardware), human labor (annotation workers, content moderators), and personal data (used without meaningful consent). The 'intelligence' in AI is the accumulated, often unpaid work of millions of people.

Argues that regulating 'AI' as a category is the wrong approach — we should regulate specific high-risk applications (predictive policing, hiring algorithms, benefits eligibility). Capability-based regulation (focusing on how powerful a model is) serves large labs by making compliance expensive.