AI: innovation for sovereignty

Considerations around Sovereign AI range from design to the infrastructure needed to create national or European channels to data security.

Such a challenge has now moved to the top of the agenda for European heads of state. Emmanuel Macron promised last May, at the Viva Technology trade fair in Paris, that France “had the capacity to host the new data centers and to develop new ones,” since this was an issue of national sovereignty. Translation: France, like its European partners, cannot take the risk of entrusting the data collected by AI to foreign companies because this collection represents a strategic issue for the future of Europe.

This desire is now at the heart of technological, political and ethical issues at this dawn of the second revolution in the digital sector, an inflexion point linked to the emergence of generative AI applications. “Until recently, AI was a subject reserved for the arcane world of digital specialists. Then ChatGPT happened… And the ease and the astonishing speed of uptake by the general public made it a social issue, virtually overnight,” summarises Thomas Le Goft, lecturer in digital law and regulation at Paris Télécom and contributor to the Report of the inter-ministerial commission on generative AI published last March. “As soon as people started asking questions like ‘how do you make a bomb?’ the authorities saw the danger.”

The European Union responded by establishing the AI Act in April 2024.

The response from European, public and private actors

This regulation aims to promote an ethical and human-centred AI, enabling the continent to stand out from American and Chinese frameworks, and to impose its own specific values and principles. “The Americans have a so-called vertical approach, where they leave it to each of their individual ministries or agencies to define rules which will aim to protect the consumer. The Chinese are following somewhat the same logic by promoting the notion that technology must promote socialist values. Europe, for its part, is seeking to promote ethical innovation,” continues Thomas Le Goft . He nevertheless notes a change in attitude among Europeans as local players capable of competing with the American giants finally begin to emerge. “From 2023, Germany, Italy and France began to actively promote home-grown innovation and research as soon as they saw the birth of unicorns capable of competing with the Americans or the Chinese. Europeans have been focussing more on the risks of generative AI but they also need to promote the development of their own AI ecosystem.”

Among these companies is Scaleway, heir to Online, the shared hosting site created by Xavier Niel in 1999. In 2023, Scaleway is investing 200 million euros in infrastructure to develop artificial intelligence services and develop the largest centre of computing power on the European continent. Today, the company is “one of the leading suppliers of cloud computing and European AI whose role is to oft er a sovereign alternative to the dominant players in the market,” explains Constance Morales, Product Marketing Manager AI at Scaleway. For the private sector, the question of Sovereign AI is central since it defines the rules of competition in a globalised economy.

For Constance Morales, this is a form of public good that reassures both market players and users. “This framework makes it possible to guarantee the implementation of regional laws and rules as is the case, for example, with the GDPR (the general data protection regulation) which guarantees the protection of the privacy of European citizens by meeting more demanding criteria than an international framework, which is off en less precise. This allows us to avoid dependence on foreign technologies and therefore reduce the risks linked to geopolitical tensions.”

The sinews of war

But given the lead taken by Silicon Valley giants like Google, Microsoff , Meta and their massive investment in generative AI in recent years, catching up may be difficult, if not impossible, for European companies. “These players, through their near-monopoly in the world of technology, have a big head start and short-cuts at their disposal to promote their own products and services. Think about how many companies are using Microsoff Oft ice and discovering new built-in AI features every day…,” Morales continues.

In the columns of the independent information site, The Conversation , Richard Florida, a doctor of urban planning at Columbia University, believes that AI will benefit two regions of the world above all others: the Bay Area of San Francisco in the USA and the Shenzhen Economic Zone of China. He thinks it is unlikely, at present, that other regions will jump on the AI bandwagon unless there is massive investment from both the private and public sectors. Because money, in support of the regulatory framework established by Europe, remains the “sinew of war” and the primary condition for an eft ective and robust Sovereign AI framework. And the member-states, alone or within European financing funds such as Digital Europe or Horizon Europe, are far from being able to compete today with the investments made in the United States and the dizzying amounts announced each week.

As an example, on 4 October, OpenAI announced a new record fundraising of $6.6 billion for the development of ChatGPT, which raised its valuation to a record $157 billion, a level of financial firepower that is far beyond the reach of European players. “Can we compete with the Americans in terms of fundraising?”, Constance Morales questions, already knowing the answer. “No, just compare the latest fundraising eft orts in Europe and the United States. It is also interesting to see how many European companies have to rely on foreign investors to raise comparable funds. Take Mistral AI, which initially received $108 million in funding from European investors but quickly turned to an American fund to raise an additional $385 million in 2023.”

So this is the paradox: European AI companies and ecosystems risk having difficulty developing without the contribution of foreign capital. Last May, Microsoff announced an investment of 4 billion euros in the development of cloud and AI infrastructure in France. Yet it was a decision welcomed more as a sign of the country’s attractiveness for inward investment than as a sign of the health of its eft orts for technological independence.

...Solutions to fill the gap

Should we conclude that Europe’s procrastination, coupled with less inward investment potential, renders the objective of Sovereign AI development and the curbing of the market share of American or Chinese companies obsolete? “Not necessarily,” says Rodolphe Renac, associate director of Alcimed, a consultancy specialising in innovation and new markets. “It is also interesting to see how our countries will position themselves because we are far from being poor at AI. France in particular is very well placed in Europe thanks to its skills in applied mathematics.”

To support his assertion, Rodolphe Renac mentions a study by Stanford University which listed the number of so-called “notable” artificial intelligence models, i.e. those which stand out due to actual, differentiated technological innovation. “The study identified 61 such models in the United States, 21 in Europe and 15 in China. Even if the pace is changing very quickly, this shows that Europe is no bit-player, particularly compared to China.

This allows us to put the orders of magnitude of advances in AI into perspective. Money is a crucial factor, but not the only one, as resources also need to flow to the most competent actors, whether public or private. This will allow us to compete eft ectively and see compelling technologies emerge as a consequence.”

To envisage an innovative industrial and business ecosystem that would also guarantee countries the means to defend themselves fairly and assert their own AI sovereignty, might require a form of global governance, along the lines of the International Atomic Energy Agency, created to ensure the security of nuclear installations and control the objectives of their programmes. “It is indeed a debate to be had,” confides Thomas Le Goft , “but it has not yet been transformed into a concrete project. People are thinking about creating a kind of international AI office that would have harmonisation of AI
frameworks at the core of its mission.”For the moment however, beyond the texts, the future of generative AI is subject only to the continued goodwill of companies and their particular notions of transparency. “Everything depends on the willingness of market players to self-regulate,” analyses Thomas Le Goft , who believes that another counterforce could emerge. I think civil society has a big role to play in this issue of Sovereign AI.”