Data vs. Warning: A Corporate Contradiction
Anthropic’s internal research and public messaging are currently at odds regarding the impact of artificial intelligence on the U.S. labor market. Peter McCrory, the company’s head of economics, released a comprehensive analysis this week arguing that AI has not yet caused a material rise in unemployment. This conclusion stands in sharp contrast to the repeated warnings from Anthropic CEO Dario Amodei, who has frequently predicted an imminent “white-collar bloodbath.”
McCrory’s analysis, which synthesizes 18 months of internal company research and Bureau of Labor Statistics data, asserts that there is no relative deterioration in employment among workers in roles highly exposed to AI automation compared to those in less-exposed positions. As of June 2026, the U.S. unemployment rate remains at 4.2%, a figure consistent with full employment, according to Federal Reserve benchmarks.
The “Jagged” Reality of AI Capabilities
McCrory attributes the lack of mass disruption to what he terms AI’s “stubbornly jagged” capability profile. Borrowing a concept from Wharton researcher Ethan Mollick, McCrory argues that no single role in the U.S. Department of Labor’s taxonomy is fully automatable by Claude. Instead, the technology currently functions as a “thought partner” rather than a direct labor substitute. High-domain expertise remains a critical factor, as experienced professionals are better equipped to oversee AI systems and correct errors, preventing wholesale displacement.
However, McCrory acknowledges specific areas of concern. He notes that hiring for entry-level roles in highly AI-exposed sectors—such as technical writing, data entry, and customer support—has softened over the past year. He describes these workers as “canaries in the coal mine,” reflecting a potential shift in labor demand that has not yet reached the broader aggregate economy.
Amodei’s Evolving Stance
CEO Dario Amodei has maintained a more alarmist position, though his rhetoric has shifted. In May 2025, he warned that AI could wipe out 50% of entry-level white-collar jobs. By early 2026, he described AI as a “general labor substitute.” Most recently, in June 2026, Amodei suggested that job loss might be an “intrinsic property” of the technology, advocating for policy interventions like universal basic income and wage insurance to mitigate the impact.
The tension lies in the timeline and scale. While Amodei views current trends as early evidence of a coming catastrophe, McCrory views the data as an adjustment period within a healthy market. Both agree that the future remains uncertain, particularly if AI reaches a stage of recursive self-improvement that could fundamentally alter the economic models currently being used to track employment.