Amid Silicon Valley's increasingly vocal predictions of AI-driven job losses, a critical problem has emerged: nobody can actually measure what's happening. Societal impacts researchers at Anthropic, the AI safety company, have begun investigating why conventional labor data—the Bureau of Labor Statistics, unemployment rates, and sectoral employment figures—fail to capture how AI is reshaping work. The disconnect is profound. While tech leaders warn of technological unemployment, official metrics show remarkably stable employment figures. This gap isn't reassuring; it suggests that existing measurement frameworks are simply blind to the types of displacement occurring. Workers losing hours, experiencing wage pressure, or transitioning into lower-wage work may never appear in unemployment statistics, which only count those actively seeking jobs. Similarly, self-employed creators, freelancers, and gig workers—precisely the populations most vulnerable to AI automation—are largely invisible in traditional employment data.