Raghuram Rajan pitches AI tax idea for govt coffers amid 'jobocalypse' fears

TEXT START: In a recent Project Syndicate column titled "How Corporations Can Mitigate an Al Jobocalypse", Rajan suggested that governments could tax the AI “tokens” companies use and use incentives to encourage businesses to retrain and retain employees.

The Dissection

The text converts a structural rupture into a manageable policy externality. It presents AI displacement as a problem of adoption speed, tax asymmetry, reskilling, and corporate reputation. Its proposed tools—AI taxes, training credits, and retention incentives—are transition-management devices designed to cushion the labor market while preserving the underlying automation race.

The article also supplies partial truth: adoption will be uneven, AI may create new demand, and fiscal policy can slow displacement or finance transfers. But those effects alter the timetable, not the destination.

The Core Fallacy

It assumes that making AI more expensive or workers more adaptable can preserve the human employment circuit. Under the Discontinuity Thesis, that circuit fails when AI becomes durably cheaper and more capable across cognitive work.

An AI tax can raise revenue, but revenue is not productive participation. Retraining only works when there remains economically necessary human work to retrain people for. The nurse-practitioner example describes task expansion, not proof that aggregate human labor demand survives once AI improves across the same frontier.

The proposal treats displacement as a temporary mismatch between workers and jobs. The deeper threat is that the jobs themselves cease to be necessary. A tax can place a toll booth beside the grave; it cannot resurrect the labor market.

Hidden Assumptions

• Governments can measure and tax AI usage, including foreign and indirect providers, without creating large avoidance channels.
• Global competition will tolerate slower or more expensive automation.
• Tax revenues will be deployed efficiently and at sufficient scale.
• Firms will retain retrained workers rather than automate the newly acquired capabilities.
• New demand and lower prices will create enough replacement employment to offset displaced labor.
• Reskilling remains valuable faster than AI absorbs the reskilled tasks.
• India’s labor-cost advantage remains meaningful as AI commoditizes software and back-office cognition.
• Corporate reputation will materially override the pressure to minimize labor costs.
• The transition is temporary rather than a progressive collapse of productive participation.

These assumptions are not demonstrated. They are the scaffolding required to make a terminal process look administratively solvable.

Social Function

Primarily transition management, with elements of partial truth and ideological anesthetic. It gives governments and corporate leaders a respectable menu of interventions while leaving ownership and control of AI capital untouched. It acknowledges disruption without confronting the terminal implication: transfers may preserve consumption, but they do not restore economic necessity, bargaining power, or productive status for the displaced majority.

The Verdict

Rajan’s proposal is a competent lag-defense memo, not a solution to the discontinuity. An AI tax may slow adoption, fund fiscal transfers, and buy institutional time. It cannot prevent P1 cognitive automation dominance, P2 coordination failure, or P3 collapse of productive participation. The article mistakes a longer fuse for a different explosive.