Why the EU will not remain the world’s digital über-regulator

EUROPE hopes to become the promised land—in the digital realm, at least. Harmful posts on social-media platforms will be rapidly removed. Texts will fly between rival messaging apps. You will be able to get apps from all over the internet, not just from your phone’s app store. And artificial intelligence (AI) models will be trained exclusively on data free of bias.

This is the noble aim of a set of new digital laws in the EU. Implementation of the Digital Services Act (DSA), which regulates social media, and the Digital Markets Act (DMA), which aims to keep big tech firms from competing unfairly, is entering a critical phase. Earlier this month, the European Commission designated what it calls “gatekeeper” firms, ranging from Alphabet to Microsoft, which will have to follow the DMA’s rules or risk vast fines.

And as if these new rules were not enough, the powers in Brussels are already negotiating their next big tech-policy package: an AI Act. By the end of the year this may impose significant changes on the AI models that power such services as Alphabet’s Bard and ChatGPT of OpenAI, a startup, which exhibit human-like skills in writing text. Not one of the ten leading models evaluated came close to being compliant with a draft passed by the European Parliament in June, according to a study by Stanford University. The proposed act includes requirements to reveal which data are used to train models and how much energy such training consumes. OpenAI’s model, called GPT-4, for instance, scored only 25 out of 48 possible points, the researchers said.

This latest wave of digital rule-making rolling out of Brussels raises one question that is particularly important: will it cement the EU’s role as über-regulator of the virtual world, established in the late 2010s with its General Data Protection Regulation (GDPR), which became the model for most of the world’s 150 privacy laws? Two new books provide a framework for predicting what may come to pass.

The first is “Digital Empires”, by Anu Bradford of Columbia Law School. In a previous tome, she noted that EU laws, particularly digital ones, tend to become global standards. She termed this the “Brussels effect”, a phrase that has caught on. Tech companies have found it cost-effective to comply with the GDPR globally. Governments around the world have adopted the law to make trade with the EU easier.

In her new book, Ms Bradford compares the EU with the other two main digital empires: America and China. Many believe that the EU will be a bystander in the clash of these two, but Ms Bradford disagrees. America’s market-driven model, which tends to place the interests of firms above those of citizens, is facing growing criticism around the world. Consequently, she thinks, America will end up being forced to adopt elements of the EU’s model. The bloc is thus likely to “remain the primary source of regulatory constraint for the tech industry”, she writes.

The authors of the second book, “Underground Empire”, paint a far gloomier picture for the old continent. Henry Farrell of Johns Hopkins University and Abraham Newman of Georgetown University also draw on their influential earlier work, a widely read paper entitled “Weaponised Interdependence”, the central thesis of which was that America uses its control over financial networks to force other countries to do its bidding. “Just as all roads once led to Rome, the world’s fibre-optic networks, financial systems and semiconductor supply chains converge on the United States, allowing it to project its might,” they argue, expanding this thesis.

One could argue that the Brussels effect describes a network that America has tended to ignore: the web of rules, mostly controlled by Europe. In the case of the DSA and the DMA that set-up may still work. Copycat laws so far are rare, but other countries are interested in drawing up their own. Firms say that implementing the newer regulations will be hard to automate. But some of the DSA’s rules, such as those for content-moderation and transparency requirements, are already being followed by American companies.

In AI, however, the chances are that the outcome will be very different. For a start, being the keeper of the rules was useful in a world based on rules and markets, but may prove irrelevant in one defined by a growing rivalry between great powers.

Secondly, in privacy and social media, the EU can rightly claim it represents one of the world’s biggest markets, so that tech giants need to follow its rules. While this is still true, Europe’s global position in AI is much weaker. According to another study by Stanford University, EU researchers have yet to contribute in a significant way to such models as GPT-4. Whereas 54% of makers of AI models were American in 2022, only 3% came from Germany, which leads the EU pack. The picture is similarly one-sided when it comes to private investment in AI: in America it amounted to $249bn between 2013 and 2022, while Germany spent only $7bn.

Finally, America seems to be adding the web of new tech rules to its own underground empire—or at least trying to neutralise Europe’s role as the rule-setter. Despite recent congressional hearings on AI, an American AI act is still unlikely. But the White House is trying to develop its own alternative to the EU’s regulatory network. In July it secured “voluntary commitments” from the principal model-makers, including OpenAI and Alphabet, to limit the technology’s risks.

Thus the geopolitical and technological “monster” that is America will eventually overpower the Brussels effect, predicts Mr Farrell. For the EU, this means that it cannot rest on its regulatory network. Instead it should redouble its efforts to strengthen its own AI industry, especially by completing the EU’s single digital market, which would make life easier for European startups. Since Europe is unlikely to become an AI superpower soon, the Centre for European Reform, a think-tank, argued in a recent report, it should also focus on getting businesses to adopt the technology.

Observers often ask how long America can use its invisible networks to throw its weight around the world. It is a good question: excessive use of network power can push other countries to seek alternatives. But perhaps the more pressing question to consider is how long the EU can lay claim to the role of global standards-setter, when it may well play such a minor role in the next wave of technology. ■