Europe’s AI Sovereignty Challenges Rooted in Dependence on U.S. Models

The European Union faces significant challenges in securing AI sovereignty as many European AI startups depend heavily on U.S.-based foundational AI models. This reliance exposes European companies and policymakers to vulnerabilities beyond the recently imposed U.S. restrictions on foreign AI access, underscoring structural dependencies in AI supply and demand in Europe’s application-layer strategy.

What Happened

Following U.S. government actions ordering Anthropic to restrict foreign access to two advanced AI models, European scrutiny of AI sovereignty intensified. The move highlighted Europe’s deep entanglement with American AI ecosystems, particularly through startups like Sweden’s Lovable that build applications on top of U.S.-developed AI models rather than developing proprietary technology. This dynamic reveals that Europe’s focus on AI at the application layer, as articulated in policy strategies such as the European Technological Sovereignty Package, faces systemic challenges due to dependency on non-European AI infrastructure and providers.

Key Facts

• The Swedish startup Lovable, valued at $6.6 billion, builds its AI-powered website and app creation tools on U.S. models from Anthropic and Google rather than proprietary tech.
• The European Commission’s AI policy emphasizes “Applied AI” at the application layer as per the 2024–2029 Commission political guidelines.
• Major U.S. hyperscalers like Google, Amazon, and Microsoft supply foundational AI models while also competing directly in application markets in Europe.
• Anthropic’s moves to integrate AI application features pose competition threats to startups that rely on its models, exemplifying “sherlocking.”
• European AI startups face increasing difficulties controlling pricing, API access, and value capture, with core infrastructure often controlled by U.S. companies.

Why It Matters

Europe’s strategy of fostering AI innovation primarily at the application layer risks perpetuating dependency on U.S.-based AI model providers, which limits true technological sovereignty. This structural dependence restricts European startups’ control over core AI infrastructure, exposes them to competitor behaviors by their own suppliers, and potentially funnels AI-generated value back to non-European companies. It also complicates efforts to establish an autonomous European AI ecosystem aligned with strategic sovereignty goals.

Background

The European Technological Sovereignty Package, introduced by Commissioner Henna Virkkunen, bundles initiatives including the Chips Act 2.0 and the Cloud and AI Development Act aimed at reducing reliance on non-European digital suppliers. The emphasis on “Applied AI” stems from political guidelines set forth by Ursula von der Leyen’s 2024–2029 Commission, promoting AI tools’ industrial adoption without competing at the cost-intensive frontier model level.

Analysis

Industry observers note that Europe’s application-layer focus mirrors a venture capital financing logic favoring asset-light startups over infrastructure-heavy investments, creating incentives for early exits to dominant U.S. tech firms. Analysts warn this dynamic may entrench the role of U.S. hyperscalers as indispensable suppliers and competitors simultaneously, eroding European control. The pattern of “sherlocking”—where providers replicate features developed by reliant startups—exemplifies this challenge, raising questions about Europe’s long-term strategic autonomy in AI.

Who Is Affected

• European AI startups, especially application-layer firms like Lovable, Harvey, and Legora.
• European policymakers aiming to enhance AI technological sovereignty.
• U.S. hyperscalers and AI model providers including Anthropic, Google, Amazon, and Microsoft.
• The broader European tech ecosystem and venture capital community.

What Remains Unclear

• How European AI policy will evolve to address the entrenched dependence on U.S. foundational AI providers beyond supply increase efforts.
• The efficacy of proposed European legislation and strategies in fostering upstream AI technology and infrastructure development.
• Whether European startups can realistically build defensible proprietary data moats against U.S. competitors in horizontal AI applications.
• The timeline and concrete measures to reduce vulnerabilities associated with control over pricing models and API access.

What Comes Next

This information was not confirmed in the reviewed sources.

Read more US News stories on Goka World News.