Data sovereignty is a first step towards AI sovereignty and AI competitiveness

On Wednesday, February 19th, the European Commission (EC) published its updated data strategy. While the strategy has three traditional “horizontal” pillars, the fourth pillar seems to draw the most attention. The goal is to proactively build “Common European data spaces in strategic sectors and domains of public interest”. These data pools have the potential to impact large parts of the European economy. As Thierry Breton, European Commissioner for Internal Market and Services states, “the industrial data war has started now, and Europe will be its main battlefield”. This strategy presents a push towards European data sovereignty. So what does it mean for AI in Europe?

We argue that indeed data sovereignty is a crucial enabler for AI in Europe and the proposed measures are a leap in the right direction. However, it is only the first step towards AI sovereignty, and AI sovereignty should not be pursued at the cost of AI competitiveness as both are critical to each other. To avoid a potentially harmful trade-off between AI sovereignty and competitiveness we suggest five key recommendations to European policy makers and business leaders that are relevant for many of the world’s large economies (e.g. Brazil, India, Canada and Japan), but particularly relevant when applied in the current European context.

Policymakers should:

1.     Consider data sovereignty together with access to crucial AI-enabling technologies

2.     Cooperate with foreign digital giants but on European terms

3.     Apply measures on a European level and support deregulation on a national level

Business leaders should:

4.     Act proactively when it comes to data sovereignty, it may well be a competitive differentiator in the near future

5.     Leverage their vertical expertise and digital capabilities, to fully embrace the new AI reality

EU data strategy presents an unprecedented push towards European data sovereignty

Data sovereignty can be defined on various levels: from individuals, to companies, to nations and regions. We define European data sovereignty as: The ability for European citizens, businesses, policymakers and other actors of society to control who accesses, processes and alters their data, and where it is stored.

The proposed European data strategy drives data sovereignty as it aims to create large datasets based on European rules, invest in European data capabilities and infrastructures, and empower individuals, and SMEs. On top of this, the creation of Common European Data spaces in strategic sectors has the potential to bring sovereignty to the currently emerging industrial data battlefield.

While we believe data sovereignty has important merits as a goal in and of itself, competitiveness in AI requires a broader perspective which needs to be taken. The whitepaper on artificial intelligence published by the EC alongside the data strategy is a good start, but it remains somewhat conceptual, with no concrete measures yet proposed.

Data sovereignty is good, but AI sovereignty is better

Data sovereignty is important, but Europe’s ambition should go further: AI sovereignty. We define AI sovereignty for Europe as access (physical or remote) to the most advanced AI-enabling technologies and data on a global level, applied according to European rules and values, without being dependent on foreign players. The triple dimensions of our definition already reveal several implications. Measures cannot result in reduced access to the most advanced technologies and data, which would isolate the economy. However, access should not be unconditional either, and European specificities should be respected. Finally, access cannot be obtained in a relationship of dependency as was the case with GPS for decades. 

Data sovereignty is a necessary condition for AI sovereignty, but not a sufficient one (see Exhibit 1). First, we consider all AI technologies and their applications, not just data-heavy ones such as the ones based on Machine Learning (ML) techniques. Second, we expand the notion towards AI-enabling technologies. If data is available, but access to AI-enabling technologies such as AI chips and supercomputers are restricted, Europe is not sovereign in AI. Finally, we add the notion of non-dependency. If Europe has access to the most advanced technologies, but needs to knock on the door of US and China for every use of it, AI sovereignty is not reached.

AI sovereignty is crucial for AI competitiveness, and vice versa

Discussions about sovereignty often invite comments about potentially negative impacts on competitiveness and indeed, it has often. It has for example been argued that Russia’s efforts to develop its own “sovereign Runet”, are harming its internet economy through loss of talent, increased bureaucratic requirements, and inefficient investments, amongst others. So is there a trade-off between AI sovereignty and AI competitiveness? And how can negative potential side-effects be avoided?

In its first wave, AI mostly focused on customer-facing applications such as search engines, and image recognition. Europe has not reached AI sovereignty so far in these technologies. But AI is on the brink of a second wave: its application by vertical industry players. If Europe continues on the current path, AI sovereignty will not be here reached either.

We argue that Europe needs to upgrade its AI competitiveness if it envisions a different outcome. We believe there are two main conditions to be AI competitive: First, it is crucial to have all AI business fundamentals in place (e.g. R&D, start-ups, talent and access to finance). Second, AI needs to be applied at scale. Where access to consumers was key for the first wave, access to industrial data and vertical industry players will now become the key differentiator. 

Without competitiveness, AI sovereignty cannot be reached. Non-competitive application results in economic loss, eventually no application altogether, no wealth creation and no sovereignty. Europe had non-dependent access to search engine technology, but lack of competitive application at scale in the Quaero debacle resulted in a permanent lack of sovereignty. Europe needs to support its vertical industry players to apply AI at scale. Efficient ecosystem competition on a European level will be crucial for competitive AI ecosystems to emerge. The Darwinian process proven successful in the Digital sectors in China[1] and the US can serve as an example.

AI sovereignty also plays a critical role in enabling AI competitiveness. European industry players will not be able to apply AI at scale if their access to enabling data and technologies remains dependent. For the first wave of AI, Europe was not ready, and this time needs to be different.

There can be no trade-off between AI sovereignty and competitiveness in Europe (see Exhibit 2). Pursuing one without pursuing the other will result in reaching neither. The time for Europe is now. With Europe’s strong industrial base, it can play an ambitious game. But to fully leverage this potential, bold action is required.

 To avoid the trade-off, five key recommendations

Policymakers should:

1.     Consider data sovereignty together with access to crucial AI-enabling technologies. Europe should strive for sovereignty in technologies such as AI chips, quantum computing and supercomputers as well. Investments need to be made in commercial applications of these technologies. Focus on research, as was primarily the case in the past, is insufficient.

2.     Cooperate with foreign digital giants but on European terms. Reaching AI sovereignty, whilst not losing AI competitiveness will require “dancing with the digital titans”. Concretely, cooperation with Chinese and US AI/platform players is inevitable if Europe is to stay competitive. However, this cooperation needs to be mutually beneficial, respecting European rules and values. Data sovereignty is a first step.

3.     Apply measures on a European level and support deregulation on a national level. This will allow the emergence of competitive AI leaders/ecosystems. Rather than adding layers of European regulation on top of existing rules, simplification on the European level should be the goal.

Business leaders should:

4.     Act proactively when it comes to data sovereignty. Developing a trust-based relationship with the data subject, ensuring its control over its own data and being transparent about how data are being used are some of the often cited methods to proactively respect data sovereignty. This is not just an act of courtesy or compliance, it may well be a competitive differentiator in the near future.

5.     Leverage their vertical expertise and digital capabilities, to fully embrace the new AI reality. The combination of both has potential to self-disrupt sectors. Finding the right balance between digital and physical is key. While internet giants consider autonomous cars as smartphones on wheels, reliability and safety will remain essential too. By finding the right AI-embracing combination of both, an AI-powered lasting competitive advantage can be developed. Failing to do so will eventually lead to being disrupted.

[1] For digital sectors, not necessarily the case for other sectors

About the authors

François Candelon is a Managing Director and Senior Partner at BCG. He is the Global Director of the BCG Henderson Institute.

Midas De Bondt is a Project Leader at BCG and an Ambassador to the BCG Henderson Institute.