Anthropic Warns: AI Slows Growth of 10 Jobs by 2034

Anthropic, the creator of the Claude chatbot, released a report in March 2026 mapping the real impact of artificial intelligence on the U.S. job market. The study introduces a metric called "observed exposure," which combines the theoretical capabilities of AI models with practical usage data, concluding that certain professions are already experiencing a slowdown in projected growth until 2034. This occurs without mass layoffs, but there are clear signs of reduced hiring for younger workers.

At the heart of the analysis is a comparison with projections from the Bureau of Labor Statistics (BLS), a U.S. government agency. According to the report, for every 10 percentage point increase in AI's task coverage, projected job growth decreases by 0.6 percentage points. The data is concrete: occupations like computer programmers have 75% of their tasks already covered by Claude in real use cases, while customer service representatives reach 70%. The impact is most pronounced on older workers, women, and higher-paid professionals.

The study examines data from over 800 professions and reveals a list of the most exposed. Programmers lead the list, followed by customer service representatives and data entry clerks, both with 67% coverage. Financial analysts and accountants also appear, with 60% and 55%, respectively. However, manual sectors like construction and agriculture remain largely untouched, with minimal exposure. According to Anthropic, this demonstrates that AI is advancing more in knowledge-based tasks but still faces legal and adoption barriers in businesses.

The effects are already visible in recent data. Since the launch of ChatGPT at the end of 2022, there has been no systematic increase in unemployment in exposed areas, according to economists Maxim Massenkoff and Peter McCrory, authors of the report. However, the hiring rate for workers aged 22 to 25 in these occupations has dropped by about 14%, suggesting that AI is closing doors for market entry. Meanwhile, the actual use of the technology is only a fraction of its theoretical potential, providing time for adaptation.

Looking ahead, the report suggests that companies like Anthropic and its rivals, such as OpenAI, need to closely monitor these patterns. Industry reactions include calls for investment in upskilling, with programs teaching how to orchestrate AI tools instead of performing routine tasks. The BLS updates its projections annually, and the upcoming 2027 data should capture more advances, potentially adjusting estimates downward if adoption accelerates.