CEPS Data Science Unit: How a Brussels Think Tank Is Mapping Europe's AI Future

How a Brussels Think Tank Is Mapping Europe's AI Future

The Centre for European Policy Studies (CEPS) operates one of the most quietly influential data science operations in European policy circles. Based in Brussels, the CEPS Data Science unit sits at the intersection of technical capability and policy translation, a position that matters more now than at any point in the past decade.

In Brief

• CEPS Data Science, led by Chief Data Scientist Pierre-Alexandre Balland, combines AI tools with policy expertise to address European governance challenges
• The unit has produced commissioned research on generative AI, foundation models, and regional AI ecosystems for EU institutions
• Recent projects include mapping Ukraine's AI competitiveness and building frameworks for a "European Ecosystem of Excellence in AI"
• The team's work on algorithmic management and hidden AI champions offers practical guidance for policymakers navigating the AI Act's implementation phase

For those tracking where Europe's AI policy actually gets shaped, the conversations happening in Vienna on May 19 at Human x AI Europe will matter.

The Institutional Architecture

CEPS, founded in 1983, functions as one of Europe's oldest independent think tanks. The Data Science unit represents a relatively recent addition to its structure, reflecting a broader institutional recognition that policy analysis without computational capacity produces incomplete maps.

Pierre-Alexandre Balland leads the unit as Chief Data Scientist. His background spans Utrecht University, MIT Media Lab, UCLA, and Harvard's Growth Lab, where he works on economic complexity and its applications to industrial policy. He also serves on the European Commission's ESIR expert group, advising on innovation policy. This dual positioning, academic rigor combined with direct policy access, shapes the unit's output.

The team includes data scientists Robert Praas and Francisco Ríos, web engineering lead Carlos Navarrete, research manager Katja Spanz, and associate research assistant Leon Oliver Wolf. Six people. The scale matters: this is not a large operation, but its placement within CEPS gives it disproportionate reach into EU policy formation.

What the Unit Actually Produces

The CEPS Data Science topic page lists the unit's recent output. Three categories dominate.

Commissioned research for EU institutions. The unit completed a study on "Generative AI and foundation models in the EU" examining uptake, opportunities, challenges, and implementation pathways. This kind of work feeds directly into how Commission officials understand the technical landscape they regulate.

Ecosystem mapping. A project titled "European Ecosystem of Excellence in AI" ran from January 2024 through October 2025, addressing what CEPS describes as the EU's "fragmented" approach to AI development. The diagnosis: Europe lacks coordination. The prescription: systematic mapping of where capabilities actually exist.

Country-specific assessments. The unit mapped AI use in priority sectors and competitiveness indicators for Ukraine, and more recently completed work on Sweden's competitiveness and investment priorities in key strategic technologies. These assessments provide granular data that aggregate EU-level discussions often miss.

The Analytical Position

The unit's published briefs reveal a consistent analytical stance. A February 2025 piece titled "Europe's hidden AI champions are vital in its quest to becoming a global tech leader" argues that the AI landscape is changing faster than headline coverage suggests. The usual big tech players no longer define the entire field.

A January 2025 brief on DeepSeek, published days after the Chinese model's release rattled markets, carried the headline "Get over the DeepSeek Panic , it might actually be a good thing." The argument: freely available, capable models expand the playing field in ways that could benefit European actors who lack the capital to compete on frontier model training.

Another January 2025 piece responded to the announcement of the USD 500 billion US Stargate Project, framing it as "Europe's wake-up call." The unit does not shy away from competitive framing, but grounds it in specific investment figures and capability gaps rather than abstract anxiety.

A December 2024 brief introduced AI World, a platform the unit co-founded to provide "key insights that are shaping the AI revolution." Balland serves as co-founder and director of this initiative, which launched in 2024.

The Algorithmic Management Question

Perhaps the most operationally relevant recent output appeared in December 2025: "Algorithmic management isn't the problem , what it optimises is."

The argument reframes a debate that has consumed significant regulatory attention. Algorithmic management (AM), the use of automated systems to direct, monitor, and evaluate workers, has drawn criticism for undermining job quality. The CEPS analysis suggests the problem lies not in the technology itself but in the objective functions these systems optimize. Build AM around efficiency alone, and worker outcomes suffer. Build it around different metrics, and the same technical architecture can improve working conditions.

This framing matters for AI Act implementation. The regulation's risk-based approach requires assessors to evaluate systems based on their deployment context and effects, not merely their technical characteristics. The CEPS analysis provides a conceptual framework for making those assessments.

The European LLM Question

A February 2024 brief called for "Launching an 'AI moonshot' to develop a European large language model." The piece argued that language models represent a category of AI technology where European absence creates strategic vulnerability.

The brief accumulated over 4,600 views, suggesting it touched a nerve. The argument connects to broader debates about digital sovereignty, data governance, and whether Europe can maintain policy autonomy over AI systems it does not build.

Eighteen months later, the question remains unresolved. European efforts to develop competitive foundation models continue, but the gap with US and Chinese capabilities has not closed. The CEPS position, that this gap represents a structural problem requiring coordinated response, has gained traction in policy circles without yet producing the "moonshot" the brief advocated.

Methodological Approach

The unit's work combines quantitative methods with policy translation. Balland's academic background in economic complexity, which uses network analysis to map how capabilities cluster and evolve, informs the unit's approach to ecosystem mapping.

Economic complexity analysis treats technological capabilities as interconnected. Countries or regions do not develop new capabilities randomly; they build on adjacent competencies. This framework helps explain why some European regions develop AI capacity while others do not, and suggests where targeted investment might produce returns.

The approach also informs the unit's work on regional divides. A policy contribution titled "Europe must act now to bridge regional divides and halt political extremism" connects economic geography to political outcomes, arguing that capability gaps between regions create conditions for political instability.

Institutional Positioning

CEPS operates multiple research units covering agriculture, energy, finance, foreign policy, migration, and other domains. The Data Science unit's cross-cutting position allows it to contribute to work across these areas while maintaining technical depth.

The unit also connects to CEPS programs including the Digital Forum and Cybersecurity@CEPS, creating internal pathways for data science perspectives to inform broader institutional output.

This positioning reflects a bet: that policy analysis increasingly requires computational capacity, and that think tanks without this capacity will produce less relevant work. The bet appears to be paying off. The unit's output receives attention from Commission officials, member state policymakers, and the broader European AI policy community.

What This Means for European AI Governance

The CEPS Data Science unit represents one model for how policy institutions can engage with AI: build internal technical capacity, produce research that bridges technical and policy audiences, and maintain independence while engaging directly with regulatory processes.

The model has limitations. Six people cannot cover the full scope of European AI policy. The unit's output necessarily focuses on selected questions, leaving others unaddressed.

But the model also has advantages. Small teams can move quickly. Direct policy access means research reaches decision-makers. Academic connections provide methodological rigor. The combination produces work that shapes how European institutions understand the AI landscape they govern.

For those tracking European AI policy formation, the CEPS Data Science unit merits attention. Not because it determines outcomes, but because it helps determine how policymakers understand the choices they face.