Atlas of AI by Kate Crawford

Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence by Kate Crawford explores the often-overlooked material, political, and environmental realities of artificial intelligence. Far from being a neutral or purely digital technology, AI is rooted in extractive industries, labor exploitation, and centralized power. Crawford connects the dots between data, labor, infrastructure, and environmental costs, offering a powerful critique of how AI is shaping our world.

Who May Benefit from the Book

• Policy makers and regulators concerned with ethical AI governance
• Environmentalists analyzing the planetary cost of modern tech
• AI and tech professionals interested in the societal impact of their work
• Academics and students in media studies, sociology, and computer science
• Activists and organizers working on digital rights and labor justice

Top 3 Key Insights

• AI relies on physical resources. It is built on global supply chains involving mining, energy, and infrastructure.
• Invisible labor powers AI. Behind automation are underpaid workers tagging data, maintaining systems, and keeping the tech running.
• Data is not neutral. AI datasets often erase context and privacy, reinforcing bias and societal inequalities.

4 More Lessons and Takeaways

• Bias is baked into AI classifications. The categories used reflect human judgment, not objective truth, often leading to discriminatory results.
• Emotion recognition lacks scientific grounding. Systems that claim to read emotions rely on outdated, unreliable psychological theories.
• Surveillance grows with AI. State and corporate use of AI blurs boundaries between security and control.
• Power is centralized. A few tech giants dominate AI development, intensifying inequality and limiting democratic oversight.

The Book in 1 Sentence

AI is not just code—it’s a global system of resource extraction, labor control, and power that shapes lives and environments.

The Book Summary in 1 Minute

Atlas of AI reveals how artificial intelligence is far from clean or objective. Kate Crawford exposes AI’s true cost, from mined minerals and carbon emissions to exploited workers and stolen data. AI systems are not only built on environmental destruction and hidden labor but also reflect deep social and political biases. Affect recognition, biased algorithms, and mass surveillance extend the reach of AI into all areas of life. Meanwhile, a few corporations centralize power and benefit the most. Crawford calls for interconnected movements for justice—linking labor rights, climate action, and digital equity—to reshape the future of AI.

The Book Summary in 7 Minutes

Kate Crawford begins Atlas of AI by breaking the myth that artificial intelligence is purely virtual or immaterial. Instead, she shows how AI is deeply tied to physical, political, and social realities.

AI Starts in the Ground

AI’s infrastructure begins with Earth’s minerals. Training models and building servers requires lithium, cobalt, and rare earth elements. These materials are extracted through environmentally harmful mining operations. For example, lithium mines in Nevada or rare earth extraction in Mongolia contribute to ecological damage. These practices often involve pollution, deforestation, and displacement of communities.

Energy consumption is another major concern. Training large AI models uses massive amounts of electricity—sometimes as much as entire cities. Much of this power comes from fossil fuels. The environmental footprint of AI is therefore comparable to aviation, yet it receives far less scrutiny.

Resource	Use in AI	Impact
Lithium	Batteries for devices & servers	Water depletion, toxic waste
Rare earths	Magnets, sensors	Soil contamination, radiation
Electricity	Powering data centers	High carbon emissions

Hidden Labor Behind Automation

Crawford challenges the narrative that AI replaces human labor. Behind every “automated” system are thousands of people—data labelers, content moderators, and warehouse workers. They perform repetitive tasks, often under exploitative conditions. For example, workers in Amazon warehouses are closely monitored and pushed to extremes.

Algorithmic management controls how workers spend every second. Surveillance systems track their movements, score performance, and reduce them to data points. Workers are pressured to meet quotas, creating a stressful and dehumanizing environment. This dynamic is present not only in physical labor but also in digital gig work.

Solidarity among workers across the AI pipeline—from miners to data annotators—is essential to push for fair wages and dignified conditions.

Data Is Not Just Data

Data is the fuel for AI. But Crawford shows how it is often gathered unethically. Facial images, social media posts, and text are scraped from the internet without consent. This erodes privacy and treats personal information as free raw material.

Turning images into datasets removes context. For instance, mugshots or selfies used in training models lose their original meaning. This leads to biased outputs. Facial recognition often performs worse for people with darker skin, creating risks like false arrests.

Ethical use of data requires consent, context, and care. Simply removing identifying details is not enough. The assumptions behind classification need scrutiny.

Classification and Bias

AI models classify people and objects based on rules made by developers. But classification is not neutral. The choices of categories reflect human values and cultural assumptions.

For example, crime prediction software may label neighborhoods as high risk based on past arrest data. This reinforces policing patterns that already target marginalized communities. Bias in the data becomes bias in the algorithm.

Fixing bias isn’t just about cleaning datasets. It requires addressing the systemic inequalities that produce the data. Otherwise, AI will continue to reproduce injustice.

The Flaws of Emotion Recognition

Crawford critiques the growing field of affect recognition, which claims to detect emotions from facial expressions. This idea traces back to psychologist Paul Ekman, who proposed that emotions are universally expressed.

But many studies challenge this claim. Emotions are shaped by culture and context. A smile might mean happiness in one setting and discomfort in another. Affect recognition ignores this complexity. Still, companies use it in hiring, policing, and education—risking misjudgments and discrimination.

AI and the Surveillance State

AI has roots in military projects. The U.S. Department of Defense and intelligence agencies have long funded AI research. Tools developed for battlefield surveillance are now used in cities, offices, and homes.

Documents from Edward Snowden’s archive show how intelligence agencies use AI to analyze massive data sets. Now, similar technologies are used by local police and private companies. The border between national security and civilian life is fading.

The “Third Offset” strategy of the U.S. military aims to keep technological dominance by partnering with tech giants. This deepens the link between Silicon Valley and military power.

The Power of Big Tech

A handful of tech companies dominate AI development. These “Great Houses of AI”—like Google, Amazon, and Microsoft—own the infrastructure, talent, and data. This centralization widens global inequality.

They also shape policy by lobbying governments and setting standards. Their interests often override public welfare. This concentration of power creates digital monopolies with little accountability.

Stronger regulation is essential to decentralize control, protect users, and encourage diverse voices in AI development.

Toward Just AI

Crawford concludes by calling for interconnected movements. AI systems reflect broader structures—capitalism, colonialism, and patriarchy. To reform AI, we must fight for climate justice, labor rights, racial equity, and data protection together.

She also introduces the “politics of refusal”—resisting the notion that AI progress is inevitable or always positive. Society must decide what kind of AI future it wants, not just accept the one built by tech elites.

About the Author

Kate Crawford is a leading scholar of artificial intelligence and its social implications. She is a Senior Principal Researcher at Microsoft Research and has held positions at MIT, NYU, and USC. Her work explores the intersection of data, politics, and power. Crawford is also a co-founder of the AI Now Institute, which focuses on the social implications of AI technologies. Through her research, she highlights the unseen costs of AI and advocates for more ethical and accountable systems.

How to Get the Best of the Book

Read slowly and reflect on each chapter. Use it as a guide to question the systems behind the AI technologies we interact with daily. Combine reading with research on real-world examples.

Conclusion

Atlas of AI is a revealing and powerful book that connects artificial intelligence to deeper issues of labor, environment, and power. Crawford invites readers to see beyond the screen—and ask what kind of world AI is helping to create.