Anthropic: Labor market impacts of AI - A new measure and early evidence

Anthropic researchers have published a groundbreaking study titled 'Labor Market Impacts of AI: A New Measure and Early Evidence,' introducing a novel methodology for tracking AI's real-world effects on employment. Unlike previous studies that relied on broad occupational categories, Anthropic's approach analyzes specific tasks within jobs using AI capability benchmarks, creating a more precise 'exposure score.' The research reveals that approximately 1.5% of US jobs are already exposed to AI automation, with technical writers, web developers, and certain financial analysts facing the highest immediate risk. This granular analysis moves beyond theoretical predictions to document early displacement occurring in the current market.

The study's task-based methodology allows for tracking how exposure evolves as AI models improve, offering a dynamic tool for policymakers and businesses. Researchers found that exposure is concentrated in white-collar, cognitive roles rather than manual labor, challenging some previous assumptions about automation. The 1.5% figure represents early evidence of impact, with the potential for significant expansion as AI capabilities advance. This work establishes a crucial baseline for measuring future displacement and highlights the need for targeted workforce adaptation strategies in high-exposure sectors.