As artificial intelligence (AI) takes on more areas of human work, critics say existing tax systems built around wage income need to change.

On May 7, IT outlet TechRadar introduced a view that, rather than imposing taxes on AI itself, tax systems should be redesigned in line with higher corporate profits and shifts in value driven by automation.

A central issue is that as AI replaces human work, the tax base collected from wages could weaken. Until now, the system has held in which individuals work, earn income and pay taxes, and governments use that revenue to provide social security and public goods. But if labour income falls and demand for public spending rises as large language models (LLMs) and AI tools spread, pressure on public finances could increase.

The discussion is also feeding into recent policy debate. RAND Corporation warned that as AI capabilities grow and spread, approaches are needed to maintain economic opportunity, social cohesion and democratic legitimacy. Sam Altman (샘 알트먼), CEO of OpenAI, previously mentioned the need for a "new social contract" in an Axios interview, saying AI superintelligence could bring major disruption.

Still, there are limits to viewing AI as a taxpayer. Taxes assume a natural person or legal entity that bears rights, duties and responsibility, but AI systems are not legal subjects that can own property, file tax returns or accept sanctions.

As alternatives, adjusting taxation of corporate profits and an automation tax are being discussed. When machines take over tasks previously done by people, income shifts from wages to corporate profits. Suggestions include calculating taxes based on the wages of replaced workers, or using indicators such as revenue per employee and the share of automated tasks within a business to reflect in taxation the extent to which human labour has shifted into capital.

Guaranteed annual income and AI dividend funds have also been mentioned as redistribution mechanisms. Under such ideas, companies that earn large profits from AI would bear the funding burden, while AI applications used in commercial environments would face a nominal compute charge to build a fund. The money raised could be distributed directly to support retraining and job transitions. An example was also cited of the Singapore government providing citizens with free premium access passes for AI courses.

The discussion does not stop at domestic tax design. Because the AI economy easily crosses borders, some also point to the need for international coordination to prevent companies from shifting locations to exploit differences in national tax rates. There are concerns that if AI taxation is not aligned globally, companies could move to low-tax regions and widen wealth gaps, making the world economy more unstable.

The focus of the debate is less on whether governments can tax AI directly than on how quickly they can reflect in institutions the value-creation structure reshaped by AI. With tax reform moving more slowly than the technological revolution, critics warn that if governments act only after wage declines become clear, they may be forced to rush policy design in a crisis. The analysis also says that if automation sharply boosts productivity while leaving many people financially unstable, redistribution could become not a choice but a tool to maintain economic stability.