A New Bill Would Tax AI Companies More the More Americans They Put Out of Work

Keypoints:

• Reps. Sara Jacobs, Greg Casar, and Valerie Foushee introduced the AI Tax and Work Protection Act
• The bill taxes AI companies on either token value or revenue, whichever is higher
• Rates start at 2% (tokens) or 3% (revenue) when unemployment is 5% or lower, and rise as unemployment climbs
• Revenue would fund a new Work Protection Administration creating jobs in infrastructure, child care, and elder care
• Anthropic CEO Dario Amodei has separately predicted AI could push unemployment to 10-20%

A new House bill ties AI companies' tax bills directly to how many Americans are out of work. Introduced by Reps. Sara Jacobs, Greg Casar, and Valerie Foushee, the AI Tax and Work Protection Act would tax large AI developers on either the value of the tokens, the small data units AI models use to process information, or on their revenue, whichever calculation produces the larger number.

The rate structure is where the bill gets genuinely novel. It starts at 2% on tokens or 3% on revenue when national unemployment sits at 5% or lower, then automatically climbs as unemployment rises, effectively turning the tax into an insurance policy that only bites harder if AI-driven job losses actually materialize at scale. 

The money raised would fund a new Work Protection Administration, a deliberate callback to New Deal-era job programs, creating jobs in areas facing labor shortages - infrastructure repair, child care, elder care, and teaching.

The bill's cosponsors point to real, if still modest, numbers behind the urgency. Employers cited AI as a contributing factor in 54,694 job losses during 2025, according to data from outplacement firm Challenger, Gray & Christmas, out of more than 1.1 million total job cuts that year. 

Anthropic CEO Dario Amodei has staked out a far more alarming public forecast separately, predicting AI could eventually push unemployment to somewhere between 10% and 20%, a range not seen since the Great Depression. 

Microsoft cofounder Bill Gates lent the idea unexpected momentum too, publicly calling for a tax on AI tokens and robots, arguing the current tax system "nudges you toward replacing people with machines" since payroll taxes apply to human workers but automation often doesn't carry an equivalent cost.

Not everyone agrees the urgency is warranted. An analysis from Reason.com argued the bill taxes "job losses that haven't happened," drawing a comparison to 1960s-era fears about computerization that ultimately didn't materialize at the scale predicted. 

The Tax Foundation raised a narrower, more technical objection, warning that any token tax needs to clearly separate consumer-facing AI use from business-to-business AI use, since taxing the latter risks slowing the very productivity gains AI is supposed to deliver. The bill has been referred to two House committees, Education and Workforce, and Ways and Means, where its actual path forward remains genuinely uncertain.