Anthropic economist refutes CEO Amodei's AI job bloodbath

Anthropic’s head of economics, Peter McCrory, published a comprehensive rebuttal this week to CEO Dario Amodei’s repeated warnings of a massive white-collar job crisis. Drawing on 18 months of internal research, McCrory argued that AI has caused no measurable rise in US unemployment. The public divergence highlights a critical debate over the speed and scale of AI-driven labor disruption that directly affects corporate workforce planning.

McCrory’s analysis points to a labor market that remains robust by traditional macroeconomic metrics. The US unemployment rate stood at 4.2% in June, a level the Federal Reserve considers full employment, while prime-age employment sat near multi-decade highs. Most importantly, updated Bureau of Labor Statistics data shows no relative deterioration in unemployment for workers in roles highly exposed to Claude’s automation capabilities compared to less-exposed peers. “I don’t expect unemployment to be noticeably higher a year from now—at least not because of AI,” McCrory wrote.

McCrory attributes this resilience to AI’s “stubbornly jagged” capability profile. No occupation in the Labor Department’s taxonomy has all its tasks fully handled by Claude, meaning complex work still requires human oversight to direct the system and correct errors. Internal evidence suggests Claude functions more as a "thought partner" that amplifies the output of domain experts, rather than a wholesale substitute for human labor.

However, McCrory’s data does not entirely clear the threat Amodei has flagged. Both executives agree that early-career workers in highly exposed roles face the greatest immediate vulnerability. McCrory conceded that hiring has already softened for young workers in AI-exposed fields, a trend aligned with Stanford research and BLS projections showing slower growth through 2034 for technical writers, data entry workers, and customer support representatives.

The internal disagreement reflects a broader tension in forecasting AI’s macroeconomic impact. Amodei has oscillated between warning of a "general labor substitute" creating a permanent underclass and adopting the Jevons paradox, where automating 90% of a task expands the remaining 10% to multiply productivity. McCrory’s findings lend more weight to the multiplier theory in the near term, though he acknowledged that if AI achieves recursive self-improvement, standard economic models could break down.

For market participants, the takeaway is that an aggregate labor market collapse remains unfounded in current data, but targeted disruption at the entry level is already materializing. Companies planning AI rollouts should expect near-term productivity gains through augmentation rather than immediate headcount reduction across the board.