A Democratic Deficit is Killing Europe’s Digital Industrial Policy, but No One Talks About It

William Burns is a fellow at Tech Policy Press.

Despite sovereigntist rhetoric about Europe needing to fire up its own high-tech credentials in a way that could rival the United States and China, the fact is that the continent already has a sizable public AI and information technology industrial complex, achieved through government subsidization of private firms and well-funded science and technology policy over many years. The results are already in—and we can see what the defects look like.

The underlying problem has rarely been a lack of plans, or a failure to coordinate and supersize them, or that they were too cautious or picked the wrong firms and technologies, but rather, inadequate levels of democratic accountability and control. The European policy debate regularly misdiagnoses this point. Instead, there is a pervasive contradiction between ends and means. How could “trustworthy, human-centric” versions of AI and other digital devices possibly emerge from policy settings that are minimally transparent and prioritize elite decision-making? Current knowledge of the frailties of human nature and the social construction of technology implies the answer would be a resounding “no.”

It should be apparent that the public and private sectors have to both be loaded down with democratic controls if we are to get the technology we want. The questions to fight about are therefore not bigger or smaller industrial policy, or which industrial policy, but how to reconfigure existing policies so that they shape technology and society in tandem towards a progressive future. These questions, which form a consistent but unimplemented socialist strand in European science and industrial policy—starting with Édith Cresson’s work as European Commissioner for science in the 1990s to the Commission’s LAB-FAB-APP report chaired by Pascal Lamy in 2017—are more important than ever as a counterbalance to authoritarianism.

At the most basic level, governance has to become a microcosm of the society we are building, such as a legally defined steering role for labor and civil society on the boards of all relevant public programs, as well as expanded funds to support organizations fulfilling these functions. Unless such a shift occurs, little will change. Politicians will perform the same rituals as before—while, of course, gas-lighting everyone, perhaps even including themselves, that the policies were new—and with yet more of the same unsatisfactory outcomes.

Democratic deficits

The European Commission currently pins its digital industrial strategy on what has been variously labeled “CERN of AI,” “AI factories,” and “AI gigafactories” (the former evoking a multinational particle physics laboratory in Switzerland, the latter, Elon Musk’s automotive plants). The terms boil down to combinations of scientific prestige and massed equipment. Messaging discipline, with slight variations, was in lockstep through official announcements such as the Competitiveness Compass in January, as well as the more recent AI Continent Action Plan this month. According to the Competitiveness Compass, the Commission will build upon “Europe’s existing world-class network of EuroHPC supercomputers” to provide “better coordination and support” for the “digitalization of public services and the integration of AI in the public sector” and “enhance competitiveness.”

EuroHPC, known legally as a “joint undertaking” because it is controlled “jointly” by the member states and the Commission, is a network of supercomputers across the continent with a combined annual operating budget of about one billion Euros. There are two other similar structures seemingly important to European industrial strategy, namely, Chips and Smart Networks and Services, intended to support, respectively, electronic components and wireless infrastructure. Chips is a merger of older undertakings dating from 2007. The supercomputer joint undertaking that Commission President Ursula von der Leyen now wants to expand was legally established in 2018. All three undertakings have been an exclusive mechanism of information exchange between private firms, a trusted inner circle of university collaborators, and the Commission. The governing regulations do not foresee wider public involvement.

Where public accountability rests is therefore incidental and generally ex-post, in other words, audit and budgetary “discharge” procedure in the European Parliament. These are mostly focused on detecting fraud and rank incompetence rather than shaping strategic intent. The official forum where work might get its tires properly kicked is the Parliamentary Science and Technology Options Assessment Panel (STOA), but it is underfunded, legally toothless, and dominated by the center-right European People’s Party (EPP). The panel had some brushes with scandal in 2022, while an effective Green Party member, Michèle Rivasi, passed away prematurely in 2023. On the progressive side of politics, the panel’s deputy chairperson, Spanish socialist Lina Gálvez, sits across an enormous brief covering the entirety of science and technology. Overall, this means we can get fewer insights than we need into the EU’s science and industrial programs as they play out, even with the best will in the world.

Pooling digital assets is another complex proposition for legal and technical reasons. Issues of intellectual property, trade secrets, and asset ownership are substantial, but they are poorly specified and lack transparency. Legal experts have focused on providing services for individual inventors, universities, and firms wanting to monopolize patents, but much less thought was given to designing contracts that foregrounded public accountability and equitable societal benefits. From this perspective, shifting the technology agenda from a focus on basic science into socially useful products is, therefore, unmapped legal territory – one of the ”unknown unknowns” never admitted in official publications.

Governing structures

Let us now break down the governance of the crucial supercomputer undertaking and see how it works. Its board includes among its “public members” government representatives from the member states and other European countries, including Turkey. Israeli officials have been members since 2023, a significant point as it is the only non-European state in the mix. At the time, the Europeans talked about accessing “Israel's expertise in technology and innovation,” but what actually played out is unclear. The “private members” include firms as well as public research institutions and universities.

The structure is controlled by a “governing board” in which the Commission wields 50% of the vote (the rest of the votes come from government representatives). Representatives from the private sector have direct membership in an “Industrial and Scientific Advisory Board,” although not in the governing board. The undertaking was established as an intergovernmental partnership between officials and the European Commission, with motives in academic science. I spoke by phone with two people with significant knowledge of the undertaking. My contacts insisted that private members only advise, never decide.

Finance to buy new supercomputers within the joint undertaking comes from the eight billion Euros “Digital Europe Programme.” As women make up more than 50% of the European population, mainstreaming gender equality in any given program is, in my opinion, a good shorthand measure of its social democratic credentials. In this case, the Commission assesses its own investment as having a “not yet clear positive impact on gender equality,” a result that ought to ring alarm bells but passed without much remark. Indeed, the overriding metric seems to be how the new computers rank in a global table of calculation speeds—likely to be of peripheral interest to society as a whole. The Dutch group Waag Futurelab has explored democratic procurement procedures, and these kinds of measures ought to be mainstreamed.

SiPearl, a chip development program that emerged from the joint undertaking, was spun out from the main contractor, French computer firm Atos. It is the physical embodiment of Europe’s public IT industrial complex and the best the system can serve up. There is no evident representation of civil society within its governing board. Put simply, if industrial strategy reproduces what already exists, in other words, the existing bureaucratic and commercial hierarchy, the social democratic gain from public effort cannot be measured. This is risky when we are talking about civilian tech in a democratic society and the need, therefore, to build the kind of broad-based support that nourishes action beyond political cycles.

Hardware contractors

The hardware inventory built up by the joint undertaking is dominated by BullSequana systems supplied by Atos (part of the company has now been renamed Eviden). Hewlett-Packard (HP), Lenovo, and Fujitsu are other, more minor, equipment suppliers as it currently stands (of course, much of the material is ultimately traceable through Taiwanese microchip factories to miners scrabbling minerals out of the earth). If we are talking about expanding the system, we are really talking about Atos, the other suppliers, and their multifaceted relationship with European procurement. Unfortunately, we cannot easily track these phenomena as it would require us to know about the procurement decisions across public authorities and private firms—another data hole limiting our ability to know what is happening.

Backing Atos, a French-domiciled firm, would be a dramatic statement of sovereigntist intent. In light of the American and Japanese firms' involvement, there are also theoretical opportunities for a program that reaches into two of the world’s major tech ecosystems. Rather than proposing autarkic capacity, the European Commission’s goal would be to orchestrate the supply chain. It would be an intriguing puzzle, particularly if we assume President Trump continues reengineering global supply chains through his tariff policy. Working with a reputed firm like Lenovo would logically tap China’s mixed functionality of science and production if it could be done.

The above is, of course, conjecture. We have no idea what is going on and cannot know because the details of deliberation are unlikely to be made public. In an ideal world, officials would evaluate the propositions of all the firms, including their forward science and research programs, and investigate how European society could gain from shaping them. But we do not live in an ideal world. The Commission can never command large multinationals (particularly when those multinationals are owned by, or interpenetrated with, European governments). Choices will, therefore, hinge on a hidden interplay of political, technological, and financial factors. The problem is that integrated technology assessment, contract negotiation, and project management are not commonly found in European wheelhouses or, for that matter, in any wheelhouses.

Ambition meets science

As with all programs hinged on science and research, there is an inbuilt uncertainty that means political efficacy can be a million miles from real-world wins. To quote the famous physicist Richard Feynman, “For a successful technology, reality must take precedence over public relations, for Nature cannot be fooled.” The science of computing at the scales we are talking about is still in its infancy, for example, in the secure transmission of data. The would-be decision maker, therefore, faces the fact that no one truly knows what to do—which is, of course, not something that many people, let alone our political leaders, would like to admit.

The points at which governments try to turn science and research policy into industrial policy are, however, notably problematic because it is here that society starts to experience effects. Officials struggle to see beyond industrial policy’s rigid language, a language that typically descends from the American military-industrial complex and its mid-twentieth-century offshoots like Japan, South Korea, and Taiwan. But these were, in their various ways, coercive regimes that offered no benchmark at all for progressives in search of social democratic inspiration. Instead, those concerned with just and sustainable outcomes must focus on democratic control. No doubt, many policy needs would become apparent as closed mechanisms were pried open, suggesting, overall, an approach to policymaking that responds to complexity in a dynamic way.