Anthropic's Economics team released an interactive scenario explorer based on technical report Korinek et al. (2026), modeling AI's effect on US GDP, employment, and labor income through 2030 across modest, substantial, and extreme scenarios. The model treats AI as task automation and augmentation, applies it only to cognitive and knowledge workers (62% of employment), and explicitly excludes robotics.
In the extreme scenario, described as AI performing nearly all knowledge-work tasks autonomously and creating essentially no new knowledge tasks for people, US GDP in 2030 would be 32.4% higher than without AI, annual GDP growth would reach 15%, and the economy would double in size every 4.5 years. Knowledge-worker unemployment would reach 17.9%, compared to 3.9% for other workers and 11.9% economy-wide. Capital's share of GDP would rise from roughly 40% to 54.8%, while total labor income barely changes. Knowledge-worker pay would fall 11.5% below the no-AI baseline, other workers would earn 33.6% more, and average pay across all workers would rise 9.7%. Anthropic notes that coders and call-center agents may need to transition to less AI-exposed occupations such as electrician or nurse. Anthropic states the extreme scenario would likely require recursively self-improving AI adopted rapidly. Across all three scenarios, labor's income share declines, from 0.6% in the modest case to 14.8% in the extreme.
Anthropics Economics team surveyed more than 10,000 Americans on their expectations and is sharing the interactive model publicly. The model has drawn criticism, with a newsletter author and AI assistants Astra and Fable arguing it bakes in the assumption that AI is not transformational. Some external reviewers also questioned whether AI-exposed occupations would shrink at all.