America's AI Jobs Data Is Flying Blind, Congress Warns

Why It Matters

A new Congressional Research Service report reveals the federal government's employer reports labor market infrastructure may be fundamentally ill-equipped to track the workforce disruption that artificial intelligence is already unleashing, and the consequences for policy could be severe.

The United States is navigating one of the most significant technological transformations in its economic history, and it may be doing so without reliable maps. A June 2026 Congressional Research Service report quietly published last week exposes a critical vulnerability at the heart of American economic policymaking: the data systems Congress and the White House rely on to understand employment statistics were built for a different era, and they are straining under the weight of AI-driven change.

The central tension is not whether AI is reshaping labor markets: that debate is largely settled. The real crisis, the CRS report suggests, is that policymakers may not know the true scale of disruption until it is too late to respond. When lawmakers reach for job market analysis to craft workforce legislation, fund retraining programs, or defend their economic records, they are increasingly relying on instruments that cannot see what they most urgently need to see.

The Big Picture

The federal government's understanding of who is working, in what jobs, and for how much money comes primarily through a network of employer-based surveys and tax filings administered largely by the Bureau of Labor Statistics and the Census Bureau. The CRS report, authored by labor policy analyst Elizabeth Weber Handwerker, walks through this architecture in detail and identifies where it breaks down precisely when policymakers need it most.

The flagship tools are well-known to economists: the Current Employment Statistics program produces monthly payroll reports measuring job gains and losses by industry; the Quarterly Census of Employment and Wages compiles near-comprehensive employment data from unemployment insurance filings; and the Job Openings and Labor Turnover Survey tracks hiring and separations each month. Together, these programs form the backbone of workforce trends analysis in the United States.

But here is the structural problem the report lays bare: these systems measure employment by what a business produces, not by occupation or what workers actually do. That distinction matters enormously when the technology disrupting the economy targets specific tasks and job types rather than entire sectors.

The report draws a historical parallel. In the 1990s, American manufacturers began contracting out factory production jobs to employment services firms. Because the workers were now classified under the "employment services" industry rather than "manufacturing," the headline labor market data obscured the true scale of industrial job loss. The same dynamic is poised to repeat with AI, the report warns, except the pace of change may be far faster and the affected occupations far more varied.

The Classification Problem

The Occupational Employment and Wage Statistics program does track employment by job type, but it operates on a three-year rolling data collection cycle, meaning rapid occupational shifts can only be detected years after they begin. Compounding this, new types of work can only be formally counted once they are added to the Standard Occupational Classification system, which is updated infrequently. "Data Scientists," for example, were not added to the SOC until 2018; before that, their employment was lumped into a catch-all category. The next SOC update is not scheduled for use until 2028.

The report notes that some observers have called for more frequent SOC updates and for adding occupational questions to the monthly JOLTS survey, which would allow policymakers to track AI-driven hiring and separation patterns in near-real time rather than waiting years for comprehensive occupational data.

On the AI adoption measurement front, there is some recent progress. The Census Bureau added questions to its Business Trends and Outlook Survey in 2025, asking businesses whether they are using generative AI and whether that use is increasing or decreasing their overall employment. Regional Federal Reserve banks have sponsored additional surveys on expected future AI use. But these efforts remain relatively new, and their longitudinal value is still developing.

When Employers Lie

Perhaps the most politically charged section of the report addresses the reliability of employer self-reporting. The CRS analysis identifies two competing incentives that can distort the data in opposite directions.

In 2025, employers cited AI in announcing layoffs of thousands of workers. The report notes that employers may have reasons to overstate AI's role in workforce reductions by using the technology as cover for cuts that might otherwise draw more scrutiny or regulatory attention. At the same time, employers may have reasons to understate AI-driven displacement, like reducing employee resistance to new technology or to avoid potential taxes on AI adoption, a proposal circulating in some policy circles.

The result is a measurement environment in which the most politically sensitive data point may be systematically distorted in either direction depending on an employer's strategic interests.

The report also surfaces a historical precedent that should give lawmakers pause. In the early 2000s, BLS asked employers involved in mass layoffs to report the cause. Those employers self-reported approximately 11,000 jobs per year being lost to offshoring. Later, comprehensive analyses examining actual trade flows found that increased Chinese imports were associated with roughly 76,000 U.S. manufacturing job losses per year during the same period, nearly seven times the employer-reported figure. The implication for AI-era job market analysis is stark: voluntary employer disclosure may dramatically undercount systemic workforce disruption.

Political Stakes

For the Trump administration, which has made strong employment numbers a centerpiece of its economic messaging, the report presents an uncomfortable possibility: the headline figures may be masking occupation-level disruption that existing payroll reports simply cannot detect. An economy that looks healthy at the industry level may be hollowing out specific job categories in ways that won't show up in official workforce trends data for years.

The administration's deregulatory posture toward AI development creates an additional tension. Pursuing light-touch AI oversight while simultaneously cutting budgets at federal statistical agencies, a direction consistent with broader DOGE-driven austerity, would compound the measurement gaps the report identifies. The CRS analysis implicitly argues for expanding data collection at the BLS and the Census Bureau, a direction that runs against current budget priorities.

For congressional Democrats, the report offers analytical ammunition for arguments that AI-driven displacement is being systematically undercounted and that workers in vulnerable occupations lack adequate early warning or policy protection. The measurement gaps documented here could be used to justify both greater investment in federal statistical infrastructure and more aggressive oversight of AI adoption in the private sector.

For Republicans, the report's findings cut both ways. The documentation of employer incentives to overstate AI's role in layoffs could support skepticism of corporate AI narratives, while the call for expanded government data collection sits uneasily with deregulatory instincts.

For the public (particularly workers in administrative, creative, and junior professional roles where generative AI is most rapidly being deployed), the stakes are the most direct. The freelance and self-employed workforce, the report notes, falls almost entirely outside employer-based measurement systems. Platform workers, independent contractors, and gig economy participants are largely invisible to the surveys and tax filings that Congress uses to understand employment statistics. If AI reshapes non-traditional work arrangements faster than traditional employment, the data blind spot grows larger still.

The Bottom Line

Two things emerge from this report that the public and policymakers should hold onto.

First, the federal government's labor market data infrastructure is not equipped to track AI's impact on employment in real time. The tools exist to measure broad industry-level job gains and losses, but the occupation-specific disruption that AI is driving (the elimination of particular tasks, the restructuring of junior roles, the displacement of specific skill sets) is largely invisible to the systems Congress depends on. By the time the data catches up, the policy window for intervention may have closed.

Second, what employers voluntarily report about AI's workforce effects cannot be taken at face value. The incentive structure documented in this report points in contradictory directions simultaneously, and the historical record from the offshoring era suggests that comprehensive, independent analysis consistently reveals far greater displacement than employer self-reporting captures.

The CRS report stops short of prescribing specific legislation, as is appropriate for a nonpartisan research document. But its policy suggestions are clear enough: add occupational questions to monthly labor surveys, update job classification systems more frequently, expand state-level unemployment insurance data collection, and invest in the statistical infrastructure needed to see the economy as it actually is.

Whether Congress acts on those recommendations or continues to navigate one of the most consequential labor market transformations in American history with instruments calibrated for a different age is now a political choice, not a technical one.