NY Times: The AI Token Tax - TaxProf Blog

TEXT START: When Bill Gates warned of the dangers posed by A.I. this week, he also weighed in on a somewhat obscure debate among politicians and economists about the best way to tax the technology.

The Dissection

The proposal is transition management disguised as system preservation. It accepts that AI may destroy jobs, then attempts to tax the automation and manufacture replacement employment through a Work Protection Administration. Its central maneuver is political: preserve the wage-and-work ritual after work has ceased to be economically necessary.

The “token” framing adds technical prestige without solving the structural problem. Tokens are merely a measurable input or billing unit. They are not the source of AI’s competitive advantage, nor a reliable proxy for displaced human labor.

The Core Fallacy

The proposal confuses extracting revenue from automation with reversing automation.

Under the Discontinuity Thesis, AI eliminates jobs because it becomes cheaper and more capable than human cognitive labor. A tax may slow adoption or fund compensatory programs, but it does not restore the productivity advantage of human workers. Competitive pressure will continue pushing firms toward automation. If the tax becomes punitive, firms will optimize token usage, shift computation into internal systems, use different model architectures, bundle AI into products, or relocate activity. The taxable surface changes; the displacement mechanism survives.

The deeper error is treating publicly funded jobs as equivalent to productive participation. If those jobs exist because the market no longer requires the labor, they are transfers wearing uniforms. They can preserve income and social order temporarily. They cannot repair the severed employment-to-wage-to-consumption circuit.

Hidden Assumptions

• Token volume will remain a stable, visible measure of AI usage.
• AI use will be directly connected to taxable AI revenue rather than embedded in broader products and services.
• Firms will absorb the tax instead of redesigning, relocating, or reducing taxable computation.
• Tax receipts will grow reliably as job destruction grows.
• The state can create enough socially acceptable work to replace economically displaced work.
• Human beings need paid employment specifically, rather than income, status, purpose, or access to resources.
• Political institutions can coordinate a durable human-only economic domain despite competitive pressure.
• Training and public employment can preserve the old social structure after productive necessity has collapsed.

The phrase “Americans want to contribute” is not an economic argument. It is an appeal to the moral identity of the industrial age. Desire for contribution does not generate market demand for human labor.

Social Function

Classification: transition management, ideological anesthetic, and partial truth.

The partial truth is that AI taxation could capture some rents and finance transfers, public services, or transition relief. The anesthetic is presenting state-created employment as a solution to technological displacement. The proposal offers “swimming lessons” on the Titanic, then insists the passengers must keep working while the ship sinks.

Its political function is to make replacement socially tolerable without admitting replacement has occurred. Universal income is rejected as dystopic because it exposes the loss of productive necessity. A government job program is more comforting because it preserves the fiction that the old order remains intact.

The Verdict

The AI token tax is a lag defense, not a reversal mechanism. It may buy time, redistribute automation gains, and postpone unrest. It cannot defeat P1, P2, or P3. Once AI severs the mass labor circuit, publicly financed jobs preserve consumption and ceremony—not productive participation. The proposal is hospice care for the wage system, with the hospice bill denominated in tokens.