Whose worldview shapes artificial intelligence?

The demographic reality of AI leadership reveals a profound concentration of power: roughly 10,000 people—overwhelmingly male, Western-educated, and concentrated in elite institutions across just two countries—are building systems that will govern billions. These leaders are 75-80% male, primarily of East Asian or white ethnicity, and trained at fewer than 30 universities. This isn't merely a diversity problem. It's an epistemic crisis where AI systems encoding the values, assumptions, and blind spots of less than 0.0001% of humanity are being deployed globally with universal claims about intelligence, objectivity, and progress.

This concentration matters because it's measurable in harm: facial recognition systems show 10-100 times higher error rates for people of color, healthcare AI fails on non-white populations, and hiring algorithms systematically discriminate against underrepresented groups. The worldviews embedded in AI—Western individualism, rationalist epistemology, efficiency-maximization—reflect not universal human values but the specific cultural context of their creators. Understanding who builds AI reveals not just demographic statistics but whose knowledge counts, whose problems get solved, and whose futures are being determined without consent.

The demographics paint a stark picture of exclusion

At the pinnacle of AI research and development, demographic homogeneity is extreme. Women comprise only 22% of the global AI workforce but this figure masks even sharper disparities: just 18% of authors at leading AI conferences are women, less than 14% of senior executives are female, and at the elite level of highly-cited researchers, women represent only 6% while men dominate at 94%. The pipeline metaphor often invoked to explain these numbers fails under scrutiny—women earn 35% of STEM degrees globally yet their representation drops precipitously as seniority increases, from 29% at entry levels to under 14% in executive leadership.

Racial and ethnic composition reveals even more troubling patterns, particularly in Western contexts where the major AI labs are headquartered. At Google, Black employees constitute 2.5% of the workforce; at Facebook and Microsoft, just 4%. Hispanic representation hovers around 5-6% across major companies. These figures represent 80-85% underrepresentation relative to the U.S. population. The numbers become more stark in technical and research roles specifically—Black employees make up less than 2% of technical staff at leading companies. The most visible manifestation of this exclusion: when researcher Timnit Gebru counted attendees at the 2016 NeurIPS conference, she found only six Black people among 8,500 participants.

Asian individuals, particularly those of East Asian and South Asian descent, are the notable exception to tech's diversity crisis in Western contexts, comprising roughly 34% of technical roles. However, this apparent "success" obscures important nuances: Asian workers face significant barriers to leadership advancement (the "bamboo ceiling"), are often pigeonholed in technical rather than managerial tracks, and this aggregate category erases vast differences among Asian ethnicities and experiences. Most critically, 70% of top AI researchers in the United States are foreign-born or foreign-educated, with Chinese nationals representing 50 of the 217 most influential researchers and Indian nationals contributing 14 more.