The EU needs to go beyond its AI Continent Action Plan and—like the US—develop an AI foreign policy. This is vital to allow the EU to amplify its existing AI efforts while insuring it against AI disruption in the economic and security spheres.
America is leading the global race in AI. In July 2025, its government published an AI Action Plan which recommends the country seek to “lead in international AI diplomacy and security” via “full-stack AI export packages”. This, the plan says, would solidify the dependency of America’s allies on its AI ecosystem—a strategy which rests on America’s frontier AI capabilities. Since the US already contains about three-quarters of global AI supercomputer performance and hosts leading frontier AI companies Anthropic, Google, OpenAI and xAI, it is clear that America’s AI leadership will be reinforced by its new foreign policy, and vice versa.
Europe, however, is a different story. The EU has recently announced a €20bn AI gigafactories initiative—but this, when combined with France’s own data centre ambitions, would like make up only around 2% of the world’s compute supply (the computing resources which train and run AI models). The EU also has little to show for its development of frontier models. French AI start-up Mistral remains the only EU frontier model developer, targeting a valuation of $10bn. By comparison, the US-based OpenAI model is reportedly on the verge of a $500bn valuation. Mistral also remains dependent on US chips, making it vulnerable to America’s trade policy whims.
In response to these developments, the EU adopted an AI Continent Action Plan to enhance its AI competitiveness through a more coherent internal policy. However, the plan pays little attention to how the EU’s foreign policy can actually amplify and strengthen its efforts in AI. Furthermore, while America’s investments and leadership in AI means it can approach AI foreign policy from a position of strength, the EU is—by necessity—entering from a position of weakness. It is thus an unrealistic goal for the EU to dominate the world with its nascent AI ecosystem; nor should this be a foreign policy aim. Instead, a successful EU AI foreign policy should guarantee that the EU will benefit from the security and economic advantages that AI provides, even if it cannot lead the global AI race for the foreseeable future.
Indispensable leverage
A European AI foreign policy requires a dual approach. First, it needs to identify and boost the EU’s strategic indispensability across the global AI value chain. Second, the EU must be willing to use these indispensable assets as leverage in return for other AI inputs, or to secure economic and security benefits.
Currently, the main area in which the EU has indispensability in AI is chip manufacturing equipment: Dutch company ASML controls 79% of global lithography, the technology that produces the most advanced AI chips, and restrictions on the export of ASML’s machines have limited China’s AI progress. In particular, Huawei and its production partner, Semiconductor Manufacturing International Corporation, have struggled to ramp up production of chips which are seven nanometers or smaller due to a lack of sophisticated lithography systems. This highlights how vital ASML is to China’s (and the global) AI value chain.
But chips alone are not enough. Other areas where the EU could also become indispensable are underexplored, with data being one example. While EU privacy laws have cemented the EU’s reputation for being stringent with access to personal data, non-personal and industrial data use is more open. From September 2026, the EU Data Act—which obliges data holders to enable data sharing with other businesses and consumers in the single market—will come into force. The Common European Data Spaces initiative will also facilitate data sharing and access across industries including health, manufacturing, energy and finance. Since the real value of AI is in its vertical application, such as health AI for drug discovery or applying AI in agriculture to making farming more productive, the EU’s harnessing of sector-specific data means it can develop another point of leverage across the global AI ecosystem.
A third example is industrial manufacturing. As AI expert Anton Leicht says, “Many avenues to the purported economic and social benefits of AI run through physically building things. Someone actually has to make the RNA vaccines and superweapons”. And the EU—led by Germany, France and Italy—has a manufacturing sector which plays a greater role in its economy as a share of GDP than in the US. The EU also runs a goods surplus with the US, which is mainly driven by cars, machines and chemicals; countries such as Denmark and Ireland have strong manufacturing capacities in areas like pharmaceuticals. Since innovations in AI will only increase demand for new products, such as AI-designed drugs and automated vehicles, the EU is well-positioned to turn its manufacturing capacity into an indispensable asset within the AI value chain.
AI to amplify and insure
The EU needs to channel these indispensable assets through a new AI foreign policy, which can also amplify the goals laid out in the AI Continent Action Plan. For example, the EU could leverage exports of lithography technologies to gain access to compute and amplify its current efforts in the development of frontier AI models—this would benefit both its economy and its security. But the EU can also fall back on its existing indispensability to secure itself against AI disruption; its manufacturing capacity can insure it against increased AI automation generating more economic value for non-EU firms. The EU could also leverage its assets to directly access the benefits of AI innovation, even if it cannot compete at the frontier of model development. For example, access to the EU’s manufacturing capacity could become conditional on security benefits, such as co-production arrangements for automated systems. The EU could also use this leverage to secure access to cutting-edge AI models which are being developed elsewhere, and deter the export of downgraded versions into the EU market.
However, an EU AI foreign policy will require broad consensus among member states on its available means—for example, there may be unwillingness to use lithography technologies due to the economic costs incurred for the Netherlands—as well as on the circumstances under which this would apply. Indeed, the EU needs to consider a scenario in which its member states face cut-offs in critical AI inputs, like in January 2025 when the Biden administration imposed quantitative restrictions in the exports of AI chips towards 18 EU member states. The EU’s reaction was muted, and eventually President Donald Trump rescinded this policy. But it is not unlikely that the US will again impose conditions on AI chip exports—Trump has already used this threat to make the EU water down its digital regulation. An EU AI foreign policy would help to deter, or allow the EU to retaliate to, such threats.
The EU’s AI foreign policy could also act as a lever to amplify the global effect of the EU’s AI regulation. For example, Europeans could make access to European data or to its manufacturing capacities conditional on signing the AI Code of Practice. This may prompt other jurisdictions—especially middle powers that lack their own frontier AI developers—to take up Europe’s AI regulatory approach. Depending on the circumstances and the extent to which the EU is willing to develop its indispensability, a leveraged AI foreign policy could also see the AI superpowers (for now, the US and China) agree to a minimum level of regulatory commitment as part of the global AI Summit Series—the only platform which has involved both the US and China from the beginning.
As AI technologies develop and grow, the US and China are pursuing global dominance in the arena. America is not only investing economic and intellectual capital in AI: the US AI Action Plan states that “a failure to meet [international AI] demand” would cause “these countries [America’s partners and third countries] to turn to our rivals”. As such, an AI foreign policy is not simply nice for the EU to have: Europeans need to go beyond low-level internal policy reform to secure their AI future. As AI revolutionises the security and economic spheres, the EU’s AI foreign policy must react to the weaponisation of critical AI inputs and effectively utilise European assets via internal consensus.
About the author:
Giorgos Verdi is a policy fellow with the European Power programme at the European Council on Foreign Relations. His research focuses on the implications of critical and emerging technologies for the EU’s competitiveness, economic security, and foreign policy.