Anthropic's head of economics finds no broad AI-driven job losses in U.S. labor data, challenging dire forecasts even as hiring slows for young workers in tech-exposed roles
After 18 months of analyzing Bureau of Labor Statistics data, occupation-level unemployment rates, and Anthropic's own research on how employees use its Claude AI tool, Peter McCrory, Anthropic's head of economics, has published findings that challenge some of the most urgent warnings about artificial intelligence and the U.S. job market. Contrary to predictions of a rapid white-collar employment crisis, McCrory's review shows no clear evidence that AI has triggered widespread job losses or a spike in unemployment-at least not yet.
McCrory's analysis, shared in a detailed essay, points to a U.S. unemployment rate of 4.2% as of June, a level the Federal Reserve considers consistent with full employment. Job openings remain roughly in line with the number of unemployed workers, and employment among prime-age adults is near multi-decade highs. Even in occupations where AI tools like Claude are most commonly used to automate tasks, unemployment rates have not risen faster than in less AI-exposed roles.
Instead, McCrory observes that AI's impact on the workplace is more nuanced. Workers are using AI to assist and refine their tasks rather than fully automating entire job functions. No occupation tracked by the Department of Labor has seen all its core responsibilities replaced by Claude or similar AI systems. This "jagged" capability profile, as McCrory describes it, means that AI is not yet a wholesale substitute for human labor in most white-collar jobs.
Conflicting Forecasts
These findings stand in contrast to the public warnings issued by Anthropic CEO Dario Amodei, who has repeatedly argued that AI could eliminate up to half of all entry-level white-collar jobs and drive U.S. unemployment to 10-20% within five years. Amodei has called for urgent policy responses, including universal basic income and wage insurance, to address what he sees as an inevitable labor shock. Yet, as of mid-2026, the aggregate labor data do not reflect the scale or speed of disruption he has described.
Both McCrory and Amodei agree that early-career workers in AI-exposed roles are most vulnerable. The disagreement centers on how quickly and severely the disruption will unfold. According to reporting by TheStreet, the divergence between internal data and executive forecasts highlights the uncertainty facing policymakers and employers as AI adoption accelerates.
Early Warning Signs
While McCrory's research finds no broad-based spike in unemployment, he does identify areas of concern. Hiring has slowed for young workers in roles with high exposure to AI automation, such as technical writing, data entry, and customer support. Stanford researchers have described these workers as "canaries in the coal mine," signaling where labor market stress may first appear. The Bureau of Labor Statistics projects slower employment growth through 2034 for these categories, aligning with McCrory's analysis of AI's reach.
Another emerging trend is the widening gap between workers who use AI as a core part of their workflow and those who do not. Employees who have integrated AI tools are becoming more productive, while others see little change. This productivity divide is beginning to show up in hiring and wage patterns, even if it has not yet translated into higher unemployment rates.
Economic Implications
The delayed arrival of a broad AI labor shock raises questions about where productivity gains are flowing. Companies with AI-fluent employees are increasing output without expanding headcount, boosting profits but concentrating benefits among a relatively small group of firms and workers. Most of these gains accrue to high-skill, high-income employees, potentially widening the income gap between AI users and the rest of the workforce.
If wage growth remains concentrated at the top, consumer spending by lower- and middle-income households could slow, creating a drag on broader economic growth. This dynamic may compound over time, especially if hiring and wage gains remain unevenly distributed. McCrory's findings suggest that while the window for proactive policy and retraining remains open, it may not last indefinitely.
For a broader perspective on how AI-driven changes are affecting major companies and their workforce strategies, see how ServiceNow's security business has adapted amid industry upheaval in this related analysis.
According to the Bureau of Labor Statistics, the U.S. labor force participation rate for prime-age workers (ages 25-54) reached 83.7% in June 2026, the highest level since 2002. Meanwhile, the number of job openings in professional and business services-a sector heavily exposed to AI-remained above 1.7 million, only slightly below the previous year's level. These figures suggest that, so far, AI adoption has not led to a collapse in demand for white-collar labor, though sector-specific risks persist.
AI's uneven impact on the labor market highlights the importance of understanding how automation interacts with specific job tasks rather than entire occupations. Many roles consist of a mix of routine and non-routine work, and AI is often more effective at handling the former. This means that workers who adapt by integrating AI into their daily responsibilities may see productivity and wage gains, while those who do not risk falling behind. For policymakers and employers, the challenge is to design training, education, and safety net programs that address these shifting dynamics before broader disruptions take hold.
