The core argument for a token tax is not to penalize AI, but to ensure the tax system doesn't artificially favor automation. It shifts the tax base from human labor (payroll, income taxes) to AI's productive capacity, measured in tokens, to prevent tax-incentivized job displacement.
A tax would raise the cost of AI experimentation, forcing firms to prioritize safe, efficiency-focused projects over speculative R&D. This 'known ROI bias' would hamper the discovery of transformative AI applications and entrench incumbents who can better absorb experimentation costs.
Implementing a token tax solely in the U.S. would create a price disadvantage for American AI companies. Customers would be incentivized to use foreign-domiciled API providers to avoid the tax, effectively subsidizing non-U.S. inference and harming the domestic AI industry.
A flat per-token tax is fundamentally flawed because token consumption doesn't correlate with economic value creation. The same number of tokens can be used for low-value tasks like generating spam or high-value tasks like legal analysis, making it an inequitable and inefficient tax mechanism.
The push for an AI token tax isn't limited to politicians. Tech leaders, including Mark Cuban, DuckDuckGo's CEO, and Anthropic's CEO Dario Amadei, have publicly supported or floated the idea, signaling a surprising openness within the industry to novel policy solutions for AI's societal impact.