As AI systems take on tasks once performed by human employees, governments are examining new tax frameworks to address the fiscal consequences of automation.
From drafting documents and writing code to handling customer queries, AI is rapidly displacing roles that have historically generated income tax revenue for governments worldwide.
Analysts and legal experts at Squire Patton Boggs warn that today’s efficiency gains from AI adoption could easily become tomorrow’s tax liabilities for businesses.
The three most discussed proposals in policy circles are the robot tax, the token tax, and the FLOP tax, each targeting a different point in the AI value chain.
A robot tax would address the fiscal imbalance created when companies replace human workers with automated systems, effectively removing income tax contributions from the equation.
Treating AI systems as capital assets rather than labour substitutes creates a tax-driven incentive for firms to automate, and a robot tax would eliminate this advantage and restore neutrality between human and automated labour.
Proponents argue that revenue generated from a robot tax could fund retraining programmes and unemployment support for workers displaced by automation, though critics warn it risks discouraging investment in new technology.
A token tax would take a different approach, applying a levy to the provider’s billed token cost and targeting AI usage directly, regardless of whether specific jobs are displaced.
Tokens are small units of data created by breaking down words into parts of words or punctuation, and AI providers already track and bill based on token usage, making them a measurable proxy for AI-driven business activity.
One significant challenge with a token tax is that models with less efficient architecture generate far more tokens for the same task than advanced models, meaning the tax burden would depend heavily on which models a business uses rather than the value of the work performed.
The third proposal, a FLOP tax, also called a compute tax, would impose a levy on AI model providers based on the volume of compute used to train or run AI systems, measured in floating point operations.
FLOP regulation already falls within the scope of the EU AI Act, which uses compute usage beyond a certain threshold as a proxy for model capability and triggers enhanced safety controls for general-purpose AI models that pose systemic risks.
Critics argue that taxing the very infrastructure needed for AI development is likely to be a self-defeating policy that could discourage investment and innovation across the sector.
A further complication is that compute resources, data centres, cloud networks, semiconductor manufacturers, and model developers are geographically dispersed, creating significant international interdependence and raising difficult questions about where a FLOP tax should be imposed and how it could be enforced.
Without international coordination, AI providers could relocate compute-intensive activities to lower-tax jurisdictions, potentially undermining the effectiveness of any such tax and intensifying geopolitical competition for AI investment.
Squire Patton Boggs authors David Naylor and Matthew Giles conclude that widespread AI adoption will reshape not only the future of work but also the cost base of most industries, creating fiscal challenges that businesses are not yet fully anticipating.
Enterprises currently deploying AI at scale should therefore monitor policy developments closely, as regulatory and tax frameworks around artificial intelligence are evolving rapidly and remain far from settled.