Empire of AI by Karen Hao: Summary and Review

Short answer: Empire of AI: Dreams and Nightmares in Sam Altman's OpenAI is an investigative book by veteran artificial intelligence journalist Karen Hao, published on May 20, 2025, by Penguin Press. Drawing on more than 250 interviews conducted over multiple years, the book argues that OpenAI abandoned its original public-interest charter to construct a commercial venture that mirrors historic colonial enterprises like Britain's East India Company. Hao documents the hidden labor performed by Kenyan data workers, the regional water depletion caused by infrastructure projects, and the internal decisions that transformed a research nonprofit into an aggressive commercial operator.

At Ban the Bots, our editorial coverage investigates the human, labor, and environmental costs created by commercial artificial intelligence systems. Hao's reporting provides extensive documentation of how corporate priorities shift burdens onto vulnerable workforces and communities while leadership frames the technology as an unalloyed global benefit.

Core Arguments in Empire of AI

Karen Hao argues that OpenAI operates like a modern colonial power by extracting cheap overseas labor, domestic data, and natural resources while justifying that extraction through rhetoric about universal progress. In her reporting, analyzed in MIT Technology Review, Hao traces how the organization departed from its 2015 charter. The founders initially structured the lab as a non-profit entity tasked with ensuring that artificial general intelligence (AGI) would benefit all of humanity.

The book details how the pressure to fund massive computational models drove OpenAI to accept outside investment, build commercial products, and prioritize market expansion. Readers interested in the theoretical foundations of OpenAI's charter can review our explainer on artificial general intelligence (AGI). Hao shows that commercial realities pushed ethical commitments aside as executive leadership chased competitive dominance over rival labs.

Hao draws a direct historical parallel to Britain's East India Company. As she explained in an interview with Tech Policy Press, imperial trading companies extracted agricultural goods, mineral wealth, and manual labor from colonized territories under the guise of spreading civilization. In Hao's analysis, commercial artificial intelligence companies operate under a similar structure by taking user data, contracting low-wage workers across the Global South to clean training sets, and leaving local populations with the environmental consequences.

The book separates this economic reality from abstract debates about hypothetical computer consciousness. While speculative risks dominate industry headlines, as documented in our background guide on AI doomers, Hao keeps her reporting focused on tangible human exploitation occurring in the present day.

Labor and Environmental Costs Documented by Karen Hao

Empire of AI documents how automated software depends directly on human laborers working under stressful, low-wage conditions in developing nations. Hao investigates the contractors in Kenya hired to label data, annotate images, and filter graphic material out of training corpora. These workers reviewed disturbing text, violent imagery, and abusive language for modest hourly compensation, shielding Western users from the toxic content that machine learning algorithms ingest during large-scale scraping.

The human toll extends beyond content moderation offices into ecological degradation around physical data facilities. Hao travels to Chile to report on community organizers and water activists fighting the construction of water-cooled server farms in drought-stressed regions. Large server facilities consume millions of gallons of potable water to prevent processors from overheating, competing directly with local agricultural communities and municipal drinking supplies.

Hao's Chilean field reporting demonstrates that computing power requires physical raw materials, electrical generation, and local water reserves. Readers can examine the wider systemic pressures of server construction in our explainer on the AI data center bubble. The book highlights the sharp contrast between corporate marketing campaigns celebrating clean digital progress and the physical depletion suffered by frontline towns.

Access Denial and Executive Relations at OpenAI

OpenAI leadership refused to grant Karen Hao formal access or sit for interviews during an extended period of roughly three years while she investigated OpenAI. In interviews discussing her book with Bloomberg, Hao stated that the organization chose not to address her reporting directly. Instead of answering detailed inquiries regarding labor conditions, resource usage, and internal governance, OpenAI closed its doors to her inquiries.

Hao bypassed this lack of executive access by cultivating relationships with rank-and-file engineers, operational staff, and departed executives. Her sourcing network produced more than 250 interviews with people who observed internal decisions firsthand. This reporting strategy allowed Hao to reconstruct private debates between technical researchers and corporate managers without relying on managed corporate public relations statements.

OpenAI has not published a line-by-line rebuttal to the book's findings. Rather than engaging directly with Hao's documentation of labor extraction and community displacement, the lab maintained corporate silence throughout the publication process. This refusal to participate highlights a broader industry pattern of evading independent journalistic scrutiny while presenting curated narratives to the public.

Critical Reception and Industry Honors

Empire of AI earned widespread critical recognition from major literary institutions and national publications following its release in May 2025. The volume became a New York Times Bestseller and was selected as a New York Times Notable Book. Reviewers praised Hao for grounding her critique in field reporting across multiple continents rather than relying exclusively on Silicon Valley boardrooms.

The book won the National Book Critics Circle Award for Nonfiction, one of the most prestigious literary honors in American publishing. Several national publications named it among the best releases of the year, including Smithsonian, Scientific American, and Elle. Critics highlighted the book for translating complex machine learning operations into a readable critique of modern global capital.

Alternative Development Models for Artificial Intelligence

Hao argues throughout the book that building massive, data-hungry computational architectures is a deliberate commercial choice rather than an inevitable law of technological evolution. During a broadcast discussion with Democracy Now, Hao emphasized that the race toward ever-larger parameter counts requires extracting massive labor and environmental subsidies from vulnerable populations. When corporate executives claim that massive scale is the only path forward, they obscure other viable technical options.

The author contrasts this dominant strategy with accountable development models that smaller labs and academic groups could pursue. These alternative models prioritize consent, public oversight, targeted datasets, and modest energy footprints. By documenting these options, Hao challenges the assumption that society must accept ecological destruction and exploited labor as the necessary price of software innovation.

Hao maintains that democratic institutions must decide how computing tools are designed and deployed. When technical decisions remain concentrated inside a handful of venture-backed boardrooms, corporate leaders naturally prioritize shareholder value over worker welfare. The book presents accountability as an urgent political necessity before industrial practices become permanently entrenched.

Reporting Connections to Labor and Policy Research

Empire of AI provides foundational evidence for researchers, regulators, and consumers investigating the corporate mechanisms behind modern automation. The book confirms the core patterns we document in our AI backlash coverage, showing that public dissatisfaction stems from concrete harms rather than superficial misunderstandings. When users push back against automated tools, they are responding to systems built on precarious labor contracts and regional environmental damage.

Readers evaluating technical claims can consult Hao's investigative reporting to separate marketing announcements from operational realities. To understand how corporate software choices affect workforces and natural resources in your community, read Karen Hao's empire of ai.

Facts above are drawn from reporting published in MIT Technology Review, Penguin Random House, Tech Policy Press, Bloomberg, and Democracy Now current as of July 2026.