As artificial intelligence spreads through the economy, policymakers are starting to ask a thorny question: should businesses pay tax for using it, and if so, how?
According to the National Law Review, several distinct models are now on the table. One is a "robot tax," a levy tied to automation that replaces human workers. Another is a "token tax," which would target the units of text that AI language models process and generate. A third is a tax on floating point operations, or FLOPs — the raw mathematical calculations that power AI systems.
Each approach reflects a different idea of what, exactly, is being taxed. A robot tax focuses on the labor AI displaces. A token tax follows how AI services are actually billed and consumed. A FLOP tax reaches deeper, aiming at the underlying computing work itself, regardless of the application.
The National Law Review frames these as options being explained and debated rather than settled policy. That distinction matters: the mechanics of how any such tax would be measured, reported, and enforced remain unresolved, and different models would fall on different players — from companies deploying AI to the providers running the compute.
Why it matters: how governments choose to tax AI could shape the cost of adopting it, influence which businesses automate and how fast, and determine whether the technology's economic gains are broadly shared or concentrated.