Governments Race to Tax AI While Struggling to Define It
Policymakers across the US and Europe are pushing to tax artificial intelligence to offset job losses, but they cannot agree on what to tax. Companies, lawmakers, and Parliament members have proposed various levies without settling on basic definitions.
The proposals hinge on terms that lack precision. Douglas Holtz-Eakin, president of the American Action Forum, pointed to concepts like an AI "token" or a "robot" tax as examples of this murkiness.
In technical terms, a token is a piece of data-such as a word-that a large language model processes to generate output. Using tokens as a tax basis would require governments to track computational units across global systems, a feat that remains technically and administratively unclear.
The definitional gap reflects a deeper problem: policymakers lack the technical knowledge to design workable tax mechanisms. Without clear boundaries around what constitutes taxable AI activity, enforcement becomes nearly impossible.
Countries face competing pressures. Labor unions and workers' advocates want protection from displacement. Business groups warn that AI taxes could slow development and investment. Governments want revenue without driving companies offshore.
Until officials settle on measurable, enforceable definitions, AI tax proposals will remain theoretical exercises rather than policy tools.
Your membership also unlocks:
