Brookings publishes a $60 billion-a-year proposal to cushion AI job losses

Protecting workers from AI disruption need not mean replacing everyone’s income, Brookings fellow Robert E. Litan argues in a proposal published October 30, 2026. His three-part transition-insurance plan would cost about $60 billion annually under a sustained 6% unemployment scenario. Its centerpiece would help displaced workers absorb a pay cut—but only after they find another job.

The proposal combines wage insurance, retraining assistance and tax credits for hiring entry-level workers. Litan presents them as complementary protections against different kinds of loss, not as a prediction that mass displacement is inevitable. He argues that preparing a safety net makes sense even if workers eventually adapt to AI as they have to earlier technologies.

The paycheck comes after the next job

Litan models wage insurance for workers involuntarily displaced for reasons other than cause, with at least three years at their previous job and prior earnings below $250,000. Eligibility would not depend on proving AI caused the job loss. Benefits would begin after the worker takes a new full-time or part-time position paying less.

• First year: 80% of the earnings loss.
• Second year: 60% of the earnings loss.
• Third year: 40% of the earnings loss.

The proposed benefits have no annual or lifetime cap. The design rewards a quicker return to work while allowing recipients to keep searching for a better-paying position. Litan also includes part-time reemployment so workers could receive support while looking for full-time work, rather than having to wait for a full-time offer before qualifying.

Wage insurance has a narrower U.S. precedent. A program introduced in 2002 covered qualifying trade-displaced workers over 50 who found work within 26 weeks. It replaced half their wage loss, subject to a $10,000 lifetime maximum. Litan’s new design would reach beyond trade-related displacement and replace a larger share of lost pay initially.

Retraining and a first foothold

The other two parts address workers whom wage insurance alone would not help. Litan describes retraining support through government-funded lifetime accounts and/or tax credits for employees or employers providing training. Entry-level hiring subsidies would target people who would not qualify for either wage insurance or retraining assistance. That separates support for recovering lost earnings from help getting established in work.

Litan argues that government must take the insurance role because private providers face two obstacles. Future AI displacement offers no reliable basis for pricing premiums, he writes. And losses could arrive together across workers and regions, rather than as independent events. That concentration of risk, in his view, makes private coverage unlikely.

The savings depend on faster reemployment

The cost estimates are scenarios, not measured results from the proposed program. Litan calculates net wage-insurance costs by assuming recipients return to work two months sooner. He also assumes a 15% tax rate on earnings from that earlier reemployment. Those illustrative savings and additional tax receipts reduce the gross cost of benefits.

The model counts all three years of payments and assumes the relevant unemployment rate persists indefinitely. Costs scale directly with unemployment: the 6% case is 50% more expensive than the 4% case, while 8% doubles it. Litan calls that sustained-unemployment assumption conservative, because workers and the labor market would adjust rather than remain permanently in the same conditions.