The Case for Imposing a Tax on Artificial Intelligence
Kevin O’Neill: Managing Director of the New Frontiers Initiative at the Rockefeller Foundation.
Not a day goes by without major headlines claiming that artificial intelligence is on the brink of transforming the economy. Even if the claims that “AI is the new electricity” turn out to be somewhat exaggerated, we must still prepare for a profound change. In fact, one of the strongest and most reliable mechanisms to ensure that society benefits from AI is also one of the most common: taxes.
What might an artificial intelligence tax look like in practice? Perhaps the most practical approach would involve targeting the key inputs and most visible metrics for AI development: energy, microchips, or computing time. The United States already imposes a 15% fee on sales of certain types of AI chips to China. While this is technically an export control, it illustrates how an AI input tax could effectively work. Alternatively, others have suggested changing how capital taxes are assessed to account for the economic shifts driven by AI. This could be an AI tax in spirit, but more general in form.
The structure of any AI tax will depend on what governments intend to achieve. However, one thing remains clear: the current dialogue is more established and urgent than it was when Bill Gates proposed a “robot tax” in 2017, a notion later echoed by Bernie Sanders and others.
Naturally, some observers may question why we should impose any taxes on AI at all. The answer to this question reflects two fundamental issues regarding tax systems and how AI changes the economy. First, many countries currently tax human workers more than they do their potential competitors in the labor market—AI systems. In the United States, approximately 85% of federal revenue comes from taxing individuals and their labor (through income and payroll taxes), while taxes on capital and corporate profits contribute far less. Technologies like AI benefit from preferential treatment in the form of generous deductions, lower corporate rates, and exclusions.
Secondly, economists predict that AI will lead to increasing returns on capital relative to labor, even if it does not cause unemployment. The more extreme version suggests the emergence of AI applications that can design, replicate, and manage themselves—meaning capital would perform the necessary work to manage itself. Under current tax policies, such a shift would widen inequality gaps and reduce government revenue as a share of GDP.
An AI tax could help level the playing field between humans and machines. Earlier this year, Dario Amodei, CEO of Anthropic, warned that AI could eliminate half of entry-level administrative jobs and drive unemployment to 10-20% within five years. The realization of these forecasts may partly depend on policy. Taxing labor more heavily than capital tips the scales in favor of automation, which replaces human workers rather than increasing their numbers. At the very least, we should not allow our tax system to facilitate unemployment.
Moreover, as financial forecasts become increasingly dire, an AI tax could shield public revenues from technological shocks. If there are mass job losses or a slowdown in hiring, governments that rely on income and payroll taxes could face financial crises even if new jobs emerge later that are suited for AI.
According to a more optimistic view, the right tax policies—coupled with a productivity boom driven by AI—could address structural financial problems. Wealthy countries are already struggling to fund healthcare and pensions for aging populations, while poorer nations face the opposite challenge: educating and employing large numbers of youth despite flimsy tax bases. Revenue generated from AI could be part of the solution for both groups.
Alternatively, revenue could be directed towards AI-related issues. Targeted taxes that return revenues to the sector from which they originate—such as the U.S. gasoline tax that funds highways or the TV license fee in the UK that supports the BBC—emphasize a goal of enhancing the public benefits brought by the taxed technology. An AI tax could serve a similar purpose: funding network upgrades, educational technology, worker training, open-source AI models, AI safety research, or providing mental health protections.
An AI tax might also support unemployment insurance and retraining for displaced workers or even advance broader AI policy goals. For example, it could help discourage excessive energy use, greenhouse gas emissions, “low-quality AI-generated content,” anti-competitive behavior, or encourage safer new energy production models.
Imposing a tax on AI may seem like a distant political dream. Policymakers may not want to stifle innovation or fall behind in the global AI race. Yet this reluctance may fade as public awareness matures. If “winning” in the AI realm means healthier populations, happier children, a more capable workforce, and stronger science—not just inflated models or increased wealth for corporations—then an AI tax could help achieve that victory.
Additionally, this tax is unlikely to stifle creativity. AI is not a nascent, fragile industry. It is a technology that has been around for 70 years, now backed by some of the largest companies in the world, with corporate investments exceeding $250 billion in 2024 alone. An AI tax could be structured to ensure it does not hinder national security, market competition, or research.
In any case, crises can change minds rapidly. If AI is blamed for mass unemployment or financial shocks, elected officials and policymakers across the political spectrum may be willing to act. It is better to prepare good options now than to improvise later.
“The world will change so quickly and dramatically that equally radical changes in policy will be required to distribute this wealth and empower more people to pursue the lives they want.” This is how Sam Altman, CEO of OpenAI, wrote in 2021. Altman was speculating about the development of more advanced general AI, but his perspective is already relevant: policies must keep pace with technology and anticipate change.
AI will reshape our economies and societies in one way or another. But the outcomes are not predetermined. The policies we choose will determine whether we enjoy a future where people and communities thrive. Imposing taxes on AI is not intended to punish creativity but to ensure that rewards are shared and risks are managed for the public good. The sooner we start working on this, the better prepared we will be to use AI to create the future we want.
