The Race — Building Democracy in the Age of AI. People race toward a shared finish line.
The Race
← Trends

Supporting case

A separate survey finds a large gap between AI employees’ and the public’s expectations

What happened

Anthropic’s October 16 survey report compares its employees with a weighted U.S. public sample of 10,980 people surveyed in August 2026. Employees, on average, expected AI could perform 75% of their own work by 2030; U.S. workers’ corresponding estimate was 28%. The staff sample was unweighted and came from a very different population.

The report identifies privacy, security and preferences for human involvement as potential barriers alongside technical capability. More frequent AI users in the public sample had expectations closer to employees’, but the comparison cannot show that using AI caused those beliefs. These are expectations, not evaluated job losses or reliable forecasts of 2030 capability. Primary research ↗.

For The Race, expectations can influence hiring, investment and institutional readiness even when they prove wrong. The contrast also cautions against treating frontier-firm experience as representative of smaller organizations or public services.

Dates and implementation stage

Development: 2026-08; publication 2026-10-16

Supporting sources

Related trends