Governments are beginning to ask whether artificial intelligence needs its own tax.
Proposals range from taxing AI computing power and energy consumption to levies on AI-generated profits or even charges on every AI query. This includes a US Democrat proposal to impose a levy on AI developers.
The concern is understandable. If AI displaces workers, governments risk shrinking income tax revenues while facing higher spending on retraining and welfare. But new proposals on pushing capital taxes risks a repeat of global challenges. Instead, VAT and similar indirect taxes on AI may already have well-worked solution
AI tax debate misses the threat that’s already here
A recent Brookings paper argues these proposals miss the point. Instead of creating new AI taxes, it suggests governments should strengthen the taxation of capital, since much of AI’s economic gains are likely to accrue to shareholders and owners of intellectual property rather than employees.
That may make sense as part of a US domestic tax debate. Internationally, however, it is a much harder proposition.
Capital can avoid. Consumption can’t
The international tax system has spent decades trying and failing to answer a simple question: where should multinational profits be taxed?
The answer has proved elusive. Intellectual property can be relocated. Investment capital flows across borders. Corporate structures evolve to minimise tax. Even the OECD’s ambitious Pillar One reforms have struggled to gain global acceptance.
Consumption taxes face far fewer of these problems
A business buying an AI subscription in Germany consumes that service in Germany, regardless of where the software was developed, where the servers are located or where the AI company’s profits are recognised. VAT already follows the place of consumption.
That is precisely why governments have rapidly expanded VAT and GST on cross-border digital services over the past decade. The destination principle works.
AI is another digital service
Most AI products are not fundamentally different from today’s cloud software or digital platforms. They are consumed as online services, often across borders.
The real tax challenge is therefore not inventing a new AI tax. It is determining the correct VAT treatment.
Is the supply software, consultancy, licensing or an automated digital service? Where is the customer established? Which country has taxing rights? How should autonomous AI agents making purchases or selling services be treated?
These are questions existing VAT systems must answer.
The technology challenge
Governments are far more likely to adapt existing VAT and GST rules than create entirely new AI taxes. Indeed, the IMF promoted VAT on digital services as the optimum route recently.
That means businesses need tax technology capable of recognising AI-driven transactions, determining the correct indirect tax treatment and adapting quickly as legislation evolves.
The debate over taxing AI should therefore move beyond capital versus labour.
The more practical question is whether tax systems can correctly tax AI where it is consumed.
For indirect tax, that answer already exists. The challenge is ensuring tax technology can keep pace with how AI changes the nature of digital transactions.
The EU Parliament is currently reviewing taxing options.
Tax technology must keep pace
Whatever governments eventually decide on AI taxation, businesses cannot wait.
Tax engines will need to identify AI-based transactions, apply evolving VAT and GST rules, distinguish between digital services and other supplies, and adapt rapidly as legislation develops.
The history of digital taxation suggests governments rarely invent entirely new tax systems. They usually adapt existing ones.
AI is likely to be no different.
The challenge for businesses is therefore not preparing for an entirely new “AI tax”. It is ensuring their tax technology can evolve quickly as governments update existing VAT, GST and direct tax rules for an AI-driven economy.
That is precisely why modern legislative-coded tax engines are becoming increasingly important. They allow businesses to respond to changing AI tax rules through software updates rather than costly ERP redesigns.
