Will AI increase inequality less than past automation?

Acemoglu's task-based model also derives AI's wage and inequality effects, distinct from its TFP estimate. He shows theoretically that "even when AI improves the productivity of low-skill workers in certain tasks (without creating new tasks for them), this may increase rather than reduce inequality" — productivity gains for lower-skill workers do not automatically translate into less income inequality. Empirically, he reports that "AI advances are unlikely to increase inequality as much as previous automation technologies because their impact is more equally distributed across demographic groups," but adds "there is also no evidence that AI will reduce labor income inequality." Instead, the paper predicts "AI is predicted to widen the gap between capital and labor income."

The result follows from the same task-allocation framework used for the TFP estimate: tasks are allocated between capital and labor based on comparative advantage, and when AI raises productivity on tasks still performed by low-skill labor without creating new tasks exclusively for them, the gain does not have to flow to workers' relative income position — the channel runs through how a task's value splits between capital and labor once comparative advantage shifts, not through whether a task is automated at all. The paper also flags that some AI-enabled "new tasks" may carry "negative social value (such as design of algorithms for online manipulation)," meaning the capital/labor split is not socially neutral even before inequality is measured.

This sits alongside Does AI growth inevitably shift wealth away from workers?, which reaches a similar capital-over-labor conclusion from a different, scenario-simulation model — two separate frameworks converging on a falling labor share as AI scales, though Acemoglu's version is a near-term (10-year), moderate-growth estimate rather than Anthropic's accelerating-growth scenarios. It also extends the task-level lens in Does automation raise or lower the skills that remaining work demands? by adding a capital-versus-labor income-share dimension to that paper's within-labor wage-and-employment analysis.

The "no evidence AI will reduce labor income inequality" finding states the absence of a demonstrated effect in available data, not a measured null result from a controlled study, and the paper does not quantify by how much the capital/labor gap is predicted to widen, only its direction. Read together with the modest TFP estimate in How much will AI actually boost productivity over ten years?, the implication is that even a small aggregate productivity gain from AI could still show up as a distributional shift away from labor, without requiring any large-scale automation shock.