HR-10044-119
Referred to the Committee on Education and Workforce, and in addition to the Committee on Ways and Means, for a period to be subsequently determined by the Speaker, in each case for consideration of such provisions as fall within the jurisdiction of the committee concerned.
Sponsored by Greg Casar (D-TX)
What it does
This bill would impose a federal excise tax on companies that develop or sell access to large-scale AI models ("foundation models"), calculated based on the volume of AI tokens processed or the revenue from AI services. The tax rate would rise automatically when the national unemployment rate increases, with a built-in suspension if unemployment spikes due to causes unrelated to AI (such as a pandemic or war). Revenue would flow into a dedicated trust fund used to: (1) create a new Work Protection Administration (WPA) inside the Department of Labor, and (2) fund a competitive grant program awarding money to governments, schools, nonprofits, and tribal entities to hire workers in areas like child care, health care, infrastructure, elder care, housing, and public education — with strong wage, benefits, and labor rights requirements attached to each funded job.
Who benefits
Workers displaced or at risk of displacement by AI automation, particularly in clerical, customer service, and routine cognitive roles. Grant-funded employees who would receive prevailing wages, health insurance comparable to federal employee coverage, 12 weeks of paid family leave, and paid sick leave. Children, seniors, people with disabilities, and low-income communities who would receive expanded services in child care, elder care, housing, and public health. State and local governments and nonprofits that would receive federal funding to expand public-sector employment. Labor unions, which would gain organizing rights protections in all grant-funded workplaces. Workers in rural and underserved areas targeted by local-hiring preferences. Researchers and BLS staff who would receive $20M/year to study AI's labor market effects.
Who is hurt
AI companies — including large technology firms and cloud providers — that develop or sell access to foundation models and would bear the direct tax burden. Startups and smaller AI companies that may face disproportionate compliance costs relative to revenue. Open-source AI developers who modify and redistribute publicly available model weights could be taxed if they generate revenue or reduce their own workforce. Businesses that use AI services and may face higher prices passed through from taxed providers. Consumers of AI-powered products and services who may see price increases. Companies that use AI to improve productivity without reducing headcount, who could still face tax liability under the token-based calculation. Academic and government researchers are explicitly exempted, but private-sector R&D is not.
Supporters argue
Supporters argue that AI automation poses a historically unprecedented threat to employment across a wide range of occupations, and that the companies profiting most from this disruption should help fund the transition costs borne by displaced workers. They contend the bill's sliding-scale tax rate — rising with unemployment — creates a direct, automatic link between AI-driven job loss and the resources needed to address it, similar in logic to unemployment insurance. The grant program's wage floors, health benefits, and labor rights requirements, they argue, would ensure that publicly funded jobs are genuinely good jobs, not low-wage stopgaps, addressing both displacement and the broader erosion of job quality documented in BLS data.
Opponents argue
Opponents argue that taxing AI token usage would penalize a technology that also creates jobs, raises productivity, and could generate broad economic gains — and that doing so before the labor market effects are fully understood risks stunting U.S. AI competitiveness relative to foreign rivals not subject to similar levies. They contend the tax base is technically ambiguous: the "fair market value of tokens processed" has no established market benchmark, creating significant valuation uncertainty and litigation risk. Critics also argue the grant program's extensive labor mandates — prevailing wages, union recognition triggers, and nondisplacement rules — add administrative complexity that may slow job creation and deter eligible entities from applying, undermining the bill's core purpose.
Constitutional context
The bill imposes an excise tax under Congress's Taxing and Spending Clause authority (Art. I, §8, cl. 1), which requires geographic uniformity — a requirement the bill appears to satisfy by applying the tax nationally. Because the tax is calculated on revenue and token volume rather than on unrealized gains, it does not raise the unresolved realization question left open in Moore v. United States (2024). The Origination Clause (Art. I, §7, cl. 1) requires revenue bills to originate in the House; this bill was introduced in the House, satisfying that requirement.
Checks and balances
The Executive Branch (Treasury/IRS and the new WPA within the Department of Labor) gains significant new administrative authority to define token valuation, set tax rates via unemployment triggers, and award competitive grants; Congress retains oversight through annual reporting requirements, appropriations control over the trust fund, and dual committee referral.
Historical precedent
The New Deal-era Works Progress Administration (1935) similarly used federal funds to create large-scale public employment across infrastructure, arts, and social services, though it was funded through general appropriations rather than a dedicated industry-specific tax.