Friday , May 7 2021
Home / The Economist Europe / The EU wants to become the world’s super-regulator in AI

The EU wants to become the world’s super-regulator in AI

Summary:
Apr 24th 2021MOST LAWS are local—except in the digital realm. When the European Union comes up with some new tech regulation, it can quickly spread around the world. Global companies adopt its typically strict rules for all their products and markets in order to avoid having to comply with multiple regimes. Other governments take more than one page from the EU’s rule book to help local firms compete. The textbook example for what has been dubbed the “Brussels effect”, is the EU’s General Data Protection Regulation (GDPR), which went into force in 2018 and swiftly became the global standard.Listen to this storyYour browser does not support the element.Enjoy more audio and podcasts on iOS or Android.Small wonder, then, that all eyes were on Brussels when the European Commission on April

Topics:
The Economist: Europe considers the following as important:

This could be interesting, too:

The Economist: Europe writes How the Irish Republic is making the best of Brexit

The Economist: Europe writes Ten years after Spain’s indignados protests

The Economist: Europe writes A court ruling triggers a big change in Germany’s climate policy

The Economist: Europe writes France worries about how to handle released terrorists

MOST LAWS are local—except in the digital realm. When the European Union comes up with some new tech regulation, it can quickly spread around the world. Global companies adopt its typically strict rules for all their products and markets in order to avoid having to comply with multiple regimes. Other governments take more than one page from the EU’s rule book to help local firms compete. The textbook example for what has been dubbed the “Brussels effect”, is the EU’s General Data Protection Regulation (GDPR), which went into force in 2018 and swiftly became the global standard.

Listen to this story

Enjoy more audio and podcasts on iOS or Android.

Small wonder, then, that all eyes were on Brussels when the European Commission on April 21st published proposed regulations on artificial intelligence (AI)—making it the first influential regulator to craft a big law on AI. Will these rules be as widely adopted as GDPR?

Recent years have seen an explosion of ethics guidelines in AI, in keeping with the hype surrounding the technology. Many hope it will boost economic growth, but others worry that AI could cause great harm, for instance if algorithms end up discriminating against certain groups of people. At least 175 countries, firms and other organisations have drawn up lists of ethical principles. But most of these fall short of describing how such things as “robustness” or “transparency” can be achieved inpractice, let alone how they can be backed up by enforceable laws, says Charlotte Stix of the Eindhoven University of Technology.

With little existing legislation on AI to draw on, the commission opted for a bottom-up approach. It created a 52-member “high-level expert group” to develop its proposals, collected further input via an “AI alliance” of interested parties and published a white paper on which everybody could comment online (1,250 groups and individuals did so). The result is a document of more than 100 pages with 85 articles and no fewer than nine annexes that tries both to mitigate the potential harm of AI and to maximise its opportunities—almost to a fault, as the many exceptions and exceptions to exceptions show.

Rather than regulating all applications of AI, the EU’s rules are meant to focus on the riskiest ones. Some will be banned outright, including services that use “subliminal techniques” to manipulate people. Others, such as facial recognition and credit scoring, are considered “high-risk” and so subject to strict rules on transparency and data quality. As with GDPR, penalties for violations are stiff: up to €30m ($36m) or 6% of global revenues, whichever is higher (in the case of a firm as big as Facebook, for example, that would come to more than $5bn).

Even more than usual, however, the devil is in the details. Facial recognition for the purpose of law enforcement in public places, which raises the spectre of pervasive surveillance, is prohibited, but only if it is done in real time and barring any other “substantial public interest”, such as finding missing children. All high-risk AI services must be tested for legal conformity, but this can often be done by the provider itself. And EU member states are encouraged to create regulatory “sandboxes”, in which firms can try out novel services without fear of being hit by a penalty.

Unsurprisingly, many interested parties are unhappy. Human-rights advocates criticise mushy language and loopholes. “There’s a real question-mark over whether the regulatory framework is robust enough,” says Sarah Chander of European Digital Rights. In contrast, business groups complain about the regulatory burden. The law will “limit the areas in which AI can realistically be used”, warns Benjamin Mueller of the Centre for Data Innovation, a think-tank supported by tech firms.

Much will still change in a legislative process that will take years, perhaps even longer than the four years it took GDPR to get from proposal to adoption. But even if things move faster, the EU may have a harder time setting global rules, or at least strict ones, says Anu Bradford of Columbia Law School, who has written a book about the Brussels effect. In the case of some AI applications, such as algorithms that can be perfected without needing to be trained using massive inputs of data, providers may decide that offering special versions in Europe is worth their while. And having been rather surprised by the success of GDPR, lobbyists will redouble their efforts to get their voices heard in Brussels.

Yet the fate of the “Artificial Intelligence Act” (AIA), as it could end up being called, may well be decided in America. If GDPR took the world by storm, it was partly because Congress not only failed to come up with any data-protection legislation of its own, but also did not bother to co-operate with lawmakers in Brussels. The new administration wants to do better, but so far the transatlantic rapprochement in AI and other things tech is off to a slow start. Only if both sides work together will they beat back China’s ambitions for tech supremacy and keep digital authoritarianism at bay.

This article appeared in the Europe section of the print edition under the headline "The Brussels effect"

The Economist: Europe
With a growing global circulation (now more than 1.5 million including both print* and digital) and a reputation for insightful analysis and perspective on every aspect of world events, The Economist is one of the most widely recognized and well-read current affairs publications. The paper covers politics, business, science and technology, and books and arts, concluding each week with the obituary.

Leave a Reply

Your email address will not be published. Required fields are marked *