CONCEPT
Corporate Industrial Policy
The decisions that private AI companies make about training data, language priorities, pricing, and optimization targets—decisions with the distributional force of government industrial policy, but made without democratic accountability and presented as neutral engineering choices.
When Anthropic decided which languages to prioritize in Claude's training, which tasks to optimize its capabilities for, and how to price access across different tiers, it made decisions with the same distributional consequences as a government tariff schedule or an industrial subsidy program.
Ha-Joon Chang's framework insists that the most consequential economic decisions are not made by markets but by the institutions that shape markets—and the leading AI companies are now those institutions, operating at global scale, accountable to shareholders rather than citizens, and presenting their choices as technical necessities rather than political ones. The concept of corporate industrial policy names this gap: the exercise of industrial policy power without industrial policy accountability. A government that decides to invest in semiconductor manufacturing rather than textile production is recognizable as an actor making a choice with distributive consequences. A company that decides to optimize for English-language software development before Yoruba-language agricultural extension is making a structurally identical choice—one that shapes