Taxing data: the future of tax in an age of AI

Trump’s protectionist tariffs have reignited an old economic debate: what works best—protecting national economies or opening borders to free trade? In Portugal, the recent election campaign largely ignored this issue. But the world has moved on. The most relevant debate today is not between protectionism and economic liberalism, but rather about how to tax a world where value has ceased to be physical and has become digital.

Artificial intelligence, which has quietly entered our lives, has brought with it a level of complexity that challenges the tax system itself. If it is already difficult for the average citizen to fill out an IRS form or understand VAT, how can they understand a tax system dominated by algorithms, automation and opaque decisions? The gap between what happens in the financial system and what the majority of the population understands is growing — dangerously.

We live in a data economy. If gold was the most precious commodity in the 19th century, and oil in the 20th century, today the value lies in data. Most of us give it away every day, for free, in exchange for access to seemingly “free” services. Google sells us answers, TikTok entertains us with videos, and Amazon anticipates what we want before we ask for it. But who pays taxes on these transactions?

The problem starts at the grassroots level: tax laws still operate in a physical world. According to the traditional concept of “fiscal nexus,” a company is only required to pay taxes in a country where it has a physical presence. But in cyberspace, this logic has collapsed. A Mongolian citizen can buy products or use services from dozens of foreign companies without any of them being based there. And, of course, without paying taxes there.

Economist Marko Köthenbürger proposes a new criterion: digital presence should be considered sufficient to establish a tax link. In other words, just as resource extraction companies do, big tech companies should pay taxes in the countries where they extract data. If oil companies pay taxes where they extract oil, why don’t Meta, Google or ByteDance pay taxes where they extract data?

The debate becomes more thorny when we consider what exactly should be taxed. Consider the case of cat videos shared on TikTok. The company doesn’t charge us, but it doesn’t pay us either. However, it uses this content to train image recognition algorithms, which it then licenses to third parties — companies or governments — for millions. Where is the line between “freely shared content” and “raw material for technological profit”? And how is the taxable amount determined?

What if, in the near future, governments agreed to exchange tax data for personal data? If, for example, a technology company agreed to provide detailed voter profiles in lieu of paying certain taxes? We would be facing a new type of state capture, with direct implications for democratic integrity.

All of this leads us to an uncomfortable conclusion: the current tax system is obsolete. It was designed to tax monetary flows, not data. It was designed for an industrial economy, not an algorithmic one. And, most importantly, it was designed for a world with physical borders, not digital ones.

The question is: are we moving towards a social credit model, where individual and collective value is measured by data generated, shared and processed? And, more importantly, what do the parties have to say about this? The absence of this discussion in the electoral programmes should concern us. Fiscal policy has never been neutral — and in the 21st century, it will be one of the main areas of dispute over power, privacy and sovereignty.

It's time we had this debate. Before artificial intelligence has it for us.