A landmark research paper published by Anthropic in March 2026 offers the most rigorous look yet at how artificial intelligence is reshaping the U.S. labor market. Authored by economists Maxim Massenkoff and Peter McCrory, the study introduces a new metric called observed exposure and traces its early effects across hundreds of occupations.
A Better Way to Measure AI Risk
Previous studies estimated AI's threat to jobs using theoretical capability alone. Massenkoff and McCrory go further by combining three data sources:
- Theoretical LLM capability - prior research on task automability
- Real-world usage - data from the Anthropic Economic Index tracking how Claude is used professionally
- Task definitions - the O*NET database, breaking 800+ occupations into constituent tasks
The result is observed exposure: a composite score that weights automated and work-related uses more heavily, giving a sharper picture of true displacement risk.
Key Findings
1. No Economy-Wide Job Loss Yet
Despite rapid AI adoption, the broader labor market shows no discernible disruption since ChatGPT's release. There is no systematic increase in unemployment for highly exposed workers, and aggregate job openings and wages remain stable.
2. Adoption Is Outpacing Measurable Impact
By December 2025, 35.9% of workers reported using generative AI tools, concentrated among younger, college-educated, and higher-earning employees. Yet actual AI adoption remains a small fraction of what these tools are theoretically capable of performing.
3. Entry-Level Hiring Is Slowing
The clearest early signal: a reduction of roughly half a percentage point per month in job-finding rates for people entering high-AI-exposure occupations. Hiring of younger workers in exposed fields has slowed, suggesting AI is compressing career entry points rather than displacing mid-career professionals.
4. Task Reallocation, Not Mass Displacement
The emerging pattern is adjustment through task reallocation and changes in career ladders, not broad layoffs. Workers in exposed roles are shifting what they do rather than losing jobs outright.
What This Means for the Future of Work
The research offers a nuanced message: AI has not triggered the feared recession for white-collar workers, but early stress fractures are visible at the entry level. For policymakers, educators, and business leaders, the window to prepare workers for AI-augmented roles is open, but it won't stay open indefinitely.
"The labor market has not experienced a discernible disruption since ChatGPT's release, undercutting fears that AI automation is currently eroding the demand for cognitive labor across the economy."
Massenkoff & McCrory, Anthropic (2026)
Read the full paper at anthropic.com/research/labor-market-impacts.
