The impact of AI and cross-border data regulation on international trade in digital services

GAICIL Insights

Trade statistics still think in containers.¹ The impact of AI and cross-border data regulation on international trade in digital services thinks in taps on a phone screen. Ruiqi Sun and Daniel Trefler analyze data from 2015 to 2020 on 28,000 mobile apps and find that AI increases an app's user base threefold, an effect they call massive. Their conclusion puts the causal effect of AI on international trade in mobile app services at more than 10-fold. Few trade shocks in history come with numbers like that.

The fuel is data. AI's handmaiden, in the authors' phrase, is data, including cross-border flows of personal data. Governments worried about privacy and national security have answered with thousands of conflicting regulations, ranging from guarantees of free trade in data under USMCA to outright data export bans. The headline finding: these rules bite. AI's massive effect is halved in countries with strict data flow restrictions. Autocracies impose the tightest restrictions on the export of data, and those restrictions cut their use of foreign apps by between 50% and 75%.

That turns data localization into something it rarely admits to being: a tariff. Restrictions on data flows are often presented as costless declarations of sovereignty. The evidence prices them, and the price is high. For mobile apps, the share of foreign users in total users is far higher than for manufactured goods or for web-based internet services such as e-commerce. Digital services are trade. Data rules are trade policy, whether their authors intend it or not.

The method deserves its own headline, because AI is measuring AI. The authors link mobile apps to AI by using a large language model to measure how far a firm's app deploys the AI described in its patent portfolio. Linking patents to products, they note, solves a long-standing problem in the innovation literature dating back at least to Kortum and Putnam. They also offer a way to estimate externalities that does not depend on patent citations, which matters because most firms have no patents. A research bottleneck that lasted decades yields to the very technology under study.

The canvas is wide: 35,575 apps used in 84 countries and exported from 64 countries. Mobile apps did not exist two decades ago and now dominate the lives of many. The paper gives them the analytical seriousness long reserved for steel and cars, and the picture that emerges is a sector where innovation, data and openness feed one another.

Who pays when borders close to data? The answer here is uncomfortable for governments that impose the rules: their own users, who lose AI-enhanced services, and their own developers, who lose foreign markets. Privacy and security are real concerns, and the authors acknowledge the first-order implications of data rules for privacy, national security and foreign interference in elections. The evidence suggests blunt restrictions are an expensive way to pursue those goals. Interoperable privacy frameworks and trade-agreement commitments that protect personal data while letting it move preserve far more of the gains.

The authors also point to a striking silence: academic research by international trade economists on these questions has been largely absent. The policy debate has sprinted ahead of the evidence, with thousands of rules written before anyone measured what they do. Work like this hands trade negotiators a number to put on the table.

The deepest implication concerns the geography of AI's benefits. The usual story concentrates AI in a few countries and a few firms. This paper tells a different story about distribution: the benefits travel, instantly, through apps that cross borders with a download. Restrict data flows and those benefits are cut in half. A country that walls off its data may keep its information at home, and keep the AI revolution out with it.

So here is the policy question in its sharpest form. If AI-enabled services become a growing share of world trade, then data governance becomes the trade negotiation of the century. Open flows guarded by sensible safeguards look like the price of admission to the gains from innovation. Closed flows look like a decision to watch those gains from outside.