The European Union has, by some measures, produced the world’s most ambitious regulatory framework for artificial intelligence. The EU AI Act, the first comprehensive AI regulation anywhere in the world, was adopted in 2024 and is being phased in through 2026 and beyond. Europe has also committed to investing tens of billions of euros in AI infrastructure, research and adoption through various funding instruments. The gap between these headline commitments and the actual state of European AI development — in terms of foundation models, compute infrastructure, private sector investment, and global competitive position — is the subject of an increasingly candid debate in Brussels and across European tech policy circles.
- The EU AI Act classifies AI systems by risk level; high-risk applications (medical devices, critical infrastructure, hiring tools) face the most stringent requirements.
- European AI investment totalled approximately €14 billion in 2025, compared to over €100 billion in the United States and €40 billion in China.
- The EU has no foundation model (like GPT-4 or Claude) developed by a European company at global competitive scale as of late 2026.
- The EU AI Office, established in 2024, oversees enforcement of the AI Act and coordinates member state AI policies.
- Mistral (France) is the leading European AI company by model capability, having raised over €1 billion — but remains significantly smaller than American and Chinese frontier labs.
The regulatory-first approach: 2019–2021
Europe’s initial AI strategy, published in 2019, reflected the Commission’s instinct to lead with rights-based frameworks rather than industrial policy. The focus was on trustworthy AI — defined by principles including transparency, human oversight and non-discrimination. This approach reflected genuine values and a legitimate reading of European competitive advantage: if Europe couldn’t build the biggest AI companies, it could set the rules that those companies operated under. The regulatory strategy was not wrong as a strategy; it was incomplete as a complete strategy.
2021–2024: The AI Act takes shape
The AI Act proposal (2021) and its tortuous legislative path to adoption (2024) produced a genuinely innovative regulatory framework — risk-based, sector-specific, and with enforcement provisions that have real teeth. The Act’s prohibition on certain AI uses (real-time facial recognition in public spaces, social scoring) and its requirements for high-risk AI systems (transparency, human oversight, technical documentation) address genuine concerns about AI misuse. Its critics — including leading European AI researchers — argued that the compliance burden would fall disproportionately on European companies and startups, while American and Chinese frontier labs would simply comply where needed to access the EU market and build capabilities unrestricted elsewhere.
2024–2026: The competitiveness reckoning
The Draghi report’s AI section was blunt: Europe is falling behind in the global AI competition in ways that have direct economic and strategic consequences. As explored in our productivity gap analysis, the absence of European hyperscale AI companies reflects structural features of European capital markets, regulatory environments and talent ecosystems that the AI Act does not address. The Commission’s response — the AI Factories initiative, providing access to EU-funded supercomputing for European AI developers; the AI Continent Action Plan; and the proposed AI Gigafactories for training frontier models — attempts to address the supply-side gap.
Whether these initiatives change the competitive position depends on execution speed that European institutions have not historically delivered. The window during which Europe could develop genuinely competitive foundation model capabilities is not unlimited, and the gap with American frontier labs is widening faster than European investment is closing it.
