Summary

Anthropic’s Version 1.0 scenario explorer models how different paths for AI capability and adoption could affect US GDP, jobs, wages and the division of income by 2030. Its results are conditional scenarios, not a forecast.

Anthropic’s Version 1.0 Economic Scenario Explorer models how different assumptions about AI capabilities and adoption could shape the US economy by 2030. Based on the company’s technical report Economic Scenarios for Transformative AI (Korinek et al., 2026), the interactive tool lets users vary their expectations and see possible effects on GDP, jobs, wages and workers’ share of income. These are conditional model scenarios, not predictions of what will happen.

Contents

A task-based view of the economy

The model treats jobs as bundles of tasks. AI may leave a task unchanged, help a worker do it faster or better, automate it, or create new tasks. The example on the explorer’s page is nursing: AI might help draft discharge instructions or plan a shift, automate some charting, and create work reviewing AI-generated care plans.

The framework then considers how changes to tasks across the economy affect productivity and the demand for workers. The outcomes depend not just on what AI can do, but on how extensively it is used and how quickly workers and companies adopt it. The task examples draw on the US Department of Labor’s O*NET occupational taxonomy.

Three paths, different economic effects

Anthropic presents three scenarios. In the modest case, AI’s economic impact is comparable to the internet’s: gains arrive gradually and fall within the historical range for new technologies. The substantial case assumes AI can perform half of all knowledge work by 2030, mostly autonomously, while many tasks are still done without AI. The extreme case assumes AI is more productive than people at most knowledge-work tasks, performs nearly all of them autonomously and creates essentially no new knowledge tasks for people. Anthropic says that last path would likely require recursively self-improving AI and rapid adoption.

The explorer reports 2030 US GDP of $34.1 trillion in the modest scenario, $36.3 trillion in the substantial scenario and $44.4 trillion in the extreme scenario, measured at 2025 prices. Its displayed changes are +1.6%, +8.3% and +32.4%, respectively.

The labor-market effects also differ. In the substantial scenario, the model has knowledge-worker wages essentially flat while workers in other occupations gain. In the extreme scenario, knowledge-worker wages fall by more than 10% by 2030, and unemployment rises beyond typical recessionary levels. The model also depicts more workers needing to change occupations in the more transformative scenarios, with transitions potentially taking time.

Growth and who receives it

The explorer separates the size of the economy from how its income is distributed. In the substantial scenario, it assigns 56.1% of GDP to labor and 43.9% to capital. In the extreme scenario, labor’s share falls to 45.2%, while capital’s rises to 54.8%. Anthropic’s model therefore describes a much larger economy in the extreme case alongside a smaller share going to workers overall; it also says total labor income is barely changed by 2030 in that scenario.

An August survey of 10,980 Americans found that the typical respondent’s answers implied outcomes close to the substantial-change scenario: GDP about 10% higher by 2030 than it would be without AI, with overall unemployment around 5%. Anthropic reports that around 10% of respondents held views aligned with the extreme scenario.

How to interpret the results

The explorer is a simplified model of the US economy, intended to show how different assumptions could produce different outcomes. Its Version 1.0 does not follow individual workers, so it gives only a coarse picture of the costs of displacement. Anthropic also notes that the model leaves out policy responses and aggregate-demand effects, among other forces that could shape actual outcomes.

The scenarios are useful for examining how AI capability, adoption and the creation or automation of tasks might interact. Their results depend on those assumptions, and the economy that emerges may differ materially from any of the paths shown.

Sources