Raghuram Rajan pitches AI tax and training incentives to tackle AI-driven job losses

URL SCAN: Raghuram Rajan pitches AI tax and training incentives to tackle AI-driven job losses
FIRST LINE: Raghuram Rajan proposes AI tax on tokens, tax credits for worker training as companies automate jobs

The Dissection

This text packages structural displacement as a manageable policy transition. It routes the problem into familiar instruments: taxes, retraining, retention incentives, reputation, and administrative coordination. The central question—who remains economically necessary once AI performs the work—is replaced with a softer question: how smoothly can firms migrate away from human labor?

The token tax addresses the fiscal asymmetry between taxed human employment and untaxed machine labor. The training credit subsidizes extending workers’ usefulness inside an economy adopting AI precisely to reduce dependence on them. The claim that lower prices and higher productivity will create new work supplies the standard demand-expansion escape hatch.

The Core Fallacy

It confuses taxing substitution with preventing substitution. A token levy may slow adoption, raise revenue, or fund transfers. It cannot restore the wage-to-consumption circuit once AI becomes cheaper and sufficiently capable across cognitive work. Firms will still face competitive pressure to replace human labor; the tax merely changes the price of the machine.

Training is treated as a bridge from displaced labor to newly complex work. Under Discontinuity Thesis mechanics, higher worker capability is irrelevant if AI improves faster or makes fewer workers necessary. The nurse-practitioner example demonstrates augmentation for selected workers, not mass labor absorption.

More output and lower prices can expand consumption while reducing labor requirements and concentrating ownership. Consumption may survive through transfers. Productive participation does not.

Hidden Assumptions

• AI adoption remains slow enough for retraining to outrun displacement.
• Training produces durable scarcity rather than temporarily upgraded workers competing with better AI.
• Firms will retain workers because credits and reputation outweigh the savings from elimination.
• Token use reliably measures labor replacement across different systems and workflows.
• Governments can tax foreign providers and prevent avoidance or substitution into untaxed systems.
• Competitive pressure will not force automation faster than policy can respond.
• Productivity gains create enough new labor demand to absorb displaced workers.
• India’s cost and skill advantages remain decisive after AI compresses the value of services labor.
• National tax systems can govern a transnational production technology without a race to the bottom.
• Employment remains the relevant social objective as production shifts toward capital and machine-mediated systems.

Social Function

Primary classification: transition management.

Secondary classifications: partial truth and ideological anesthetic.

This is not pure copium. It identifies a real payroll-tax asymmetry, acknowledges uneven adoption, and proposes measures that could buy time or redistribute some gains. But it anesthetizes the terminal question by treating a collapse of productive necessity as a skills mismatch. It makes post-employment disruption legible to governments and executives: tax the tokens, subsidize training, preserve reputation, and call the delay a transition.

That may manage the descent. It does not change the destination.

The Verdict

Rajan’s proposal is a plausible lag defense, not a systemic rescue. An AI tax could slow the kill rate and finance transfers; training credits could preserve selected servitor roles during the overlap period. Neither reverses P1, defeats P2, nor prevents P3. The article is a polished administrative response to discontinuity: it puts a meter on the machine, pays humans to remain useful a little longer, and mistakes delayed obsolescence for survival.