AI Access as a Geopolitical Risk: Europe’s Trade-Off Between Openness and Sovereignty

On 12 June, the Trump administration ordered Anthropic, a leading U.S. artificial intelligence (AI) company, to restrict access of its latest models to foreign nationals, citing potential ways to bypass the models’ safeguards. The company had no choice but to comply and restricted access globally, as it could not reliably distinguish American users from foreign ones. After all, employees of U.S.-based firms using its systems often have diverse, undisclosed nationalities.

EU policymakers and industry leaders quickly understood the risks: what would prevent the current U.S. president, or a future one, from restricting access to other widely used AI models? And what if future models were also unavailable to European users? Would the continent’s economy risk falling further behind?

EU leaders such as French President Emmanuel Macron, backed by stressed the need for cross-border cooperation on AI development and regulation. Conversely, several members of the European Parliament saw the decision as further evidence of U.S. unreliability, calling for greater “technological sovereignty,” increased investment in European AI capabilities and a stronger industrial policy.

While policymakers disagree on solutions, they agree on the problem: the EU’s economy, and in particular critical sectors such as technology, remains heavily reliant on third countries. The Anthropic case may seem anecdotal, but it reflects a broader trend: key technological assets, such as AI models, are concentrated in a small number of jurisdictions, meaning control over these innovations translates into geopolitical power. In other words, Europe’s dependence on foreign technology risks undermining its economic security.

Many EU officials are therefore proposing solutions similar to those used in industrial policy: simplification (AI Omnibus), economic sovereignty (Tech Sovereignty Package) and support for “EU tech champions.” The difference lies only in the partner of concern: while the EU fears industrial dependency on China, it fears technological dependence on the United States.

AI Goes Beyond Industrial Logic

However, equating AI’s role in the economy with that of traditional industry may carry risks. Industrial production can be localized, and its value chains are more tangible. By contrast, AI is a systemic and scale-driven technology that permeates the entire economy and will shape both services and industry. Its inputs are not limited to tangible resources, but also includes data, talent, computing power, and cross-border ecosystems. A failure to effectively support or regulate AI would not weaken a single sector, but risks holding back the entire economic system.

If the challenge is systemic, the response must be as well. Building European AI capabilities depends in part on physical inputs such as semiconductors, which are being addressed through initiatives like the Tech Sovereignty Package. But achieving AI leadership also requires a strong position across the broader value chain, including data, computing power, and the models used by European businesses and citizens.

This suggests that applying traditional industrial logic alone may not be sufficient. Subsidizing producers of physical goods might save some industrial output on the continent, but it will likely not secure Europe’s position in AI.

Restrictions on AI models could be approached in a similar way to trade barriers or export controls. In the latter case, policymakers typically avoid full protectionism, recognizing that complete control over value chains is neither feasible nor efficient. Rather, they seek partnerships with like-minded countries to build economic synergies, while strengthening internal competitiveness, for example through deeper Single Market integration or progress toward a Capital Markets Union. These efforts, though incomplete, support both industrial and service sectors in the short and long term, and the same logic could apply to AI.

Building AI Leadership Through Cooperation

In other words, strengthening Europe’s position in the emerging AI economy requires strategic positioning, not full control. This points toward the importance of open, cross-border technological ecosystems that allow value chains to be shaped not solely within Europe, but together with trusted global partners. It also raises questions around the potential implication of restrictive sovereignty requirements in proposals like the Cloud and AI Development Act (CADA). Ongoing efforts to conclude digital trade agreements, alongside broader trade deals, and engagement through forums such as the EU-India Trade and Technology Council or renewing the EU-U.S. Trade and Technology Council are steps in the right direction.

Businesses can also benefit from increased financial support for R&D projects. Greater support for cross-border public and private R&D initiatives, both within the EU and with like-minded partners abroad, could further strengthen these dynamics.

To build AI leadership, the EU must create an ecosystem in which forward-looking companies can thrive. This will depend on the extent to which policymakers prioritize global partnerships, support innovative R&D and enable private capital to flow more freely across borders. Anthropic’s case highlights that AI is entering a geopolitical phase. It also underscores the opportunity for the EU to strengthen its competitiveness through cooperation and open investment. More protectionist approaches, on the other hand, may carry risks in terms of cost, efficiency and access to innovation.