Transatlantic cooperation on AI and national security

Executive summary

The links between AI and national security are a key issue for testing the strength of the transatlantic relationship and assessing its prospects. US export controls on AI, while largely targeted at China, have profound implications for Europe’s ambitions to innovate, increase its competitiveness, and achieve digital sovereignty. Absent more structured cooperation, EU-US AI policy risks drifting into a form of “managed interdependence,” characterized by asymmetric leverage and ad hoc bargaining rather than strategic alignment.

There are three main areas posing barriers for greater EU-US cooperation on this topic:

• The first is the unpredictability of US policy in relation to export controls and in relation to China specifically. There seems to be disagreement within the US administration about how to treat China, and the different directions of export control policy reflect that policy trade-offs are at times being managed differently. It is not clear if export controls are primarily meant as a tool for national security, revenue raising, or industrial policy. European leaders are worried that any aspect of US policy could change in any US-China deal.
• Second, when it comes to China, the US administration is far more willing to openly treat China as a threat; but there is not yet a clear, unified EU position, with member states taking different positions.
• Third, the United States seems to give insufficient weight to the interests of US allies. This was true under the Biden administration to some degree; the second Trump administration has gone further in this regard, primarily seeing allies in transactional terms—as “export destinations”—rather than as partners with their own interests to protect and who should share in the benefits of cooperation.

In contexts such as Russia’s full-scale invasion of Ukraine, the EU and the United States have been able to cooperate closely on export controls and sanctions. Now, Europe’s priority in transatlantic trade policy talks with the United States should be demonstrating the value of similar cooperation in relation to AI and China. The EU is home to numerous world-leading firms and research hubs critical to AI. A more unified EU approach to national security and export controls could help ensure Europe can leverage these assets to increase its geopolitical and economic heft. Without more unity, the efforts of both the European Commission and individual member states to bolster and apply their respective economic-security-policy tool kit will remain constrained. European AI policy will then continue to be shaped by commitments designed to avoid higher US tariffs and export controls, including agreements to purchase large volumes of US AI chips and to invest hundreds of billions of euros in strategic sectors within the United States.

Cooperation is in the United States’ interest too. Excessive use of unilateral tools contributes to European desires to derisk from US supply chains. And cooperation offers a way to reconcile the US desire to increase exports of AI technology globally with the reality that such exports carry significant leakage risks (meaning that the technology can often find its way to places like China in circumvention of US export controls).

At the political level, achieving a grand strategic agreement on AI at the transatlantic level appears unfeasible now. The EU is not participating in the US Pax Silica initiative, through which the United States is attempting to prevent China from controlling parts of the AI supply chain, though Sweden and Greece have signed up. In the future, a “club” comprising the EU, the United States, and other like-minded democracies could be more effective at developing a collaborative industrial and economic security policy. A more pragmatic approach would be to consider what steps the United States and Europe could take to boost mutual understanding including:

• Regular (structured or informal) policy dialogue to ensure each side understands the other’s approach and to avoid surprises.
• Joint scenario planning and risk assessments to identify AI security risks, potential policy tools to mitigate them, and ways to predict and prepare responses to potential retaliation.
• Collaborative work to boost public-private engagement to ensure policies are informed by real-world capabilities and risks.

Most importantly, the EU and the US administration should identify and work on related areas where cooperation is more feasible. For example, markets for technology products like legacy chips—where the EU and the United States could cooperate to stop China from swamping global markets and driving out non-Chinese suppliers—could offer a promising avenue for transatlantic cooperation given the significant EU and US industrial interests at stake.

Context and policy background

The links between AI and national security are a key issue for testing the strength of the transatlantic relationship and assessing its prospects. The return of Donald Trump to the White House and his administration’s 2025 National Security Strategy have created unprecedented concern in Europe about the credibility of the United States as a security guarantor and the prospects for EU-US cooperation. In the field of AI, the president’s second administration’s early policy decisions—unwinding tentative steps toward AI regulation put in place by his predecessor and emphasizing unconstrained technological development in AI as key to US security—put the United States squarely at odds with Europe’s more cautious approach.

However, upon closer inspection, even under President Biden, coordination with Europe on AI and national security was limited. The first Trump administration unilaterally adopted export restrictions to keep China several generations behind in chip making, a policy that Biden continued. Washington also pressured the Netherlands to stop the export of high-end lithography machines to China, despite the opposition of ASML, the leading producer of such machines and one of Europe’s only technology success stories. As fears grew that China would develop its own lithography industry to rival the Dutch, Washington also tried to pressure European countries to impose export controls on components of ASML’s machines: controls which would target German manufacturing firms Zeiss and Trumpf and chemical producer BASF. Tension did not only relate to export controls but also to industrial policy. The US Chips and Science Act and the Inflation Reduction Act—the Biden administration’s two major industrial policy initiatives—were explicitly designed to attract foreign companies in the fields of semiconductors and green technologies to establish a presence in the United States. They were therefore criticized by allies, including the Europeans, for a lack of prior consultation.

The lowest point in transatlantic relations regarding AI probably occurred when the Biden administration introduced extensive export controls on AI chips from the United States, known as the “AI diffusion rule.” The export controls did not only target “countries of concern”: namely, US geopolitical rivals like China. It also set up a tiered approach to other countries, including allies, dividing them into aligned allies who would have minimal restrictions on access to AI chips; a larger group of countries which would face stricter licensing and allocation limits for these chips; and “countries of concern” which would face strict controls or denials on access to advanced AI technology. While many EU countries were designated as key allies and partners of the United States with a high level of security and trust standards, many EU member states were still subject to harsher rules. For example, Greece, Portugal, Poland, Latvia, Estonia, and Lithuania were subject to stricter licensing rules than other EU countries—despite their vigorous support for many US security initiatives. While the rule was supposedly reflective of whether a country had adequate means to counter smuggling of chips to places like China or Russia, EU leaders harbored concerns that the rule would effectively undermine the EU’s “single market” by precluding the free movement of goods (such as AI chips) and services (such as AI-enabled services).

The Trump administration has both elements of continuity and discontinuity with the Biden administration. On the one hand, the administration rescinded the AI diffusion rule which had irked EU governments and (at least initially) shifted toward maximizing access to US AI technologies for allies and partners, arguing that Biden’s restrictions harmed US industry. The US Bureau of Industry and Security (BIS) described the AI diffusion rule as having “stifled American innovation and saddled companies with burdensome new regulatory requirements,” without emphasizing the rule’s implications for transatlantic relations.

Continuity with the Biden administration, however, is evident when it comes to export controls against Chinese companies and the inclusion of subsidiaries of large Chinese companies in the entity list. The administration “strengthened” chip-related export controls by issuing guidance about the risks of using Chinese AI chips and about protecting supply chains. Trump also launched an investigation under Section 232 of the Trade Expansion Act of 1962 to assess whether the importation of microchips, equipment, and related products could pose a threat to national security. In April, the Trump administration strengthened export controls on China, including to prohibit the shipment to China of cutting-edge AI chips like Nvidia’s H20 chip (which had been designed to comply with the previous US export controls). In December, the United States announced its Pax Silica initiative, through which the United States is attempting to prevent China from controlling parts of the AI supply chain. It is unclear the extent to which the EU has been deliberately excluded (for example because of its commitment to regulating AI or because the United States seeks to sideline Brussels in favor of direct engagement with European capitals) and the extent to which EU countries (other than Sweden and Greece) deliberately decided not to sign onto the initiative. While Washington has been trying to persuade the Netherlands to join the initiative, for example, it has been hesitant to agree to an explicitly US-led initiative that sidelines the EU.

The Trump administration’s approach has been, overall, somewhat unpredictable, making it difficult for Europe to know where it stands. The toughening of export controls in April proved short-lived. The administration’s AI Action Plan, released in July 2025, represented a further step toward promoting the export of AI chips and computing capabilities, rather than constraining them. The plan takes the position that promoting US exports in the AI stack—and therefore encouraging other countries to become reliant on technologies the United States dominates—would best protect US national security. However, this seems at odds with the subsequent Pax Silica initiative which aims to mobilize “the complementary industrial and technological strengths of strategic companies and firms from our respective economies.” And despite Pax Silica seeking to privilege partners with “fair market practices,” Nvidia also announced that the Trump administration had reversed course and allowed the export of the H20 to China, apparently in relation to a deal related to China allowing exports of rare earth metals and reportedly conditioned on a 15 percent fee on sales.

In September 2025, the US policy direction shifted again as the administration proposed a change to its export controls (the so-called affiliates rule), which significantly increased the scope of the US controls. China reacted by imposing significant restrictions on the global export of rare earths impacting many of China’s trading partners including those in Europe. Following the Trump–Xi meeting in South Korea in October 2025, both sides agreed to a temporary de-escalation: China suspended those rare-earth export controls for one year, while the United States suspended the affiliates rule, which would have broadened controls to any entity majority-owned by a listed party. These moves aim to stabilize markets while maintaining underlying strategic competition, especially in semiconductors. In the meantime, the US president in December stated an intention to allow Nvidia to sell its H200 chips, which are close to the cutting edge, in return for a 25 percent fee on sales.

Temporarily suspending certain restrictions and export controls, both sides are aiming to stabilize markets and reduce immediate economic friction. However, strategic competition in areas such as semiconductors remains intense. The US president has also declared that Nvidia’s most advanced Blackwell chips will be reserved exclusively for US firms and will not be sold to China or other foreign buyers—illustrating a dual-track strategy of easing immediate pressures while preserving US dominance in cutting-edge technology. US export controls are increasingly pulled between three logics: national security (denial), industrial policy (advantage), and commercial diplomacy (dealmaking). A consistent theme across both the Biden and Trump administrations is the limited engagement with allies, including the EU, when shaping AI-related national-security policies. The Biden and Trump administrations reflect different views on the right trade-off between openness and security when dealing with high-end, dual-use technologies, which can be used for both commercial and military purposes. On the one hand, diffusion can encourage geopolitical rivals to rely on Western technologies, reducing their incentives to develop indigenous technology (and, to the extent they build technological ecosystems around the tech, potentially making it more difficult to divest from Western technology in future); and it can buy time to help preserve the West’s lead in cutting-edge tech. On the other hand, wider access to the technology can make it easier for other countries to master and exploit it in ways that pose security risks for the West. Policymakers must grapple with those risks as well as the strong incentives to circumvent export controls.

As the European Union currently lags in most areas of the AI stack, US policies have a significant impact on Europe. For its part, Europe is trying to balance regulation and innovation, ensuring that the two work together to make the EU competitive in AI without undercutting the values and rights of European citizens and businesses. Taking stock of these developments, this paper discusses:

• Opportunities and risks entailed in US export controls on advanced semiconductors.
• Impacts of US AI national security policies on Europe’s AI ambitions.
• Prospects for EU-US cooperation on this topic and pathways to design an effective export control policy.

Assessing the opportunities and risks of US export controls

A starting point in assessing the opportunities and risks of US export controls is making sense of competing US policy objectives:

• The plan to restrict access to advanced AI chips to US firms seems to indicate a significant expansion of export controls. In one sense, it suggests a desire to protect America’s technology lead for national security purposes.
• Yet the desire to restrict AI chips even among trustworthy US allies suggests that industrial policy considerations also are at stake. Notably, US willingness to make concessions on export controls in its “deals” with China contribute to an impression that export controls are no longer solely about national security but also about trade and geopolitical competition. It is unclear how this relates to the more collaborative Pax Silica initiative.
• The Trump administration seems to see export controls as a potential source of new public revenue, as its deals with chipmakers to grant export licenses in return for a fee.

These objectives can be understood as reflecting three partially competing logics: national security (restricting adversaries’ access to critical technologies), industrial policy (preserving and extending US technological leadership), and commercial diplomacy (using export controls as leverage in broader geopolitical and trade negotiations). The tension between these logics helps explain the apparent inconsistency in US policy. If we look at the history of export controls, the CoCom export controls against the Soviet Union during the Cold War show, for example, how they entail different types of trade-offs.

First, objectives can be in tension: e.g., national security vs. industrial policy. While export controls can increase national security, they can also negatively impact industries and, consequently, the national economic base. In 2019, for instance, Japan tightened export controls on semiconductor materials to South Korea. Japanese firms experienced delayed approvals, reduced demand, and supply chain disruptions, leading to production shifts and strategic diversification.

Second, there is the classic trade-off between security and innovation. Export controls can increase security, but they reduce openness—and therefore probably innovation, too. But being completely open could create the risk of adversaries catching up with critical technologies and perhaps using them to power their military apparatuses, but they also adversely impact exporters (both those subject to export controls and those impacted by retaliatory measures) and can deprive them of scale and revenue they need to maintain a technological edge. These trade-offs are evident in the current competition between the United States and China. Supporters of a looser policy argue that the previous approach of applying stringent export controls was ineffective. They claim it harmed the economic prospects of US exporters, prompted China to retaliate by withholding critical minerals—most prominently rare earth magnets, which reportedly influenced the administration’s reversal on H20 and H200 chip-export restrictions—which impacted many industries. The US approach to dealmaking with foreign countries on export controls may be an attempt to assist US companies to keep innovating—but the impacts of this strategy are limited unless there is consistency across different deals with third countries and predictability about the US government’s objectives in negotiations, so that companies can understand the restrictions they are likely to face.

Third, there is a trade-off between the benefits of cutting off a rival from a strategic technology to hold back their development and the benefit of keeping that rival reliant on that technology. For example, the threat (and reality) of export controls has accelerated China’s push for high-end chipmaking capabilities. Evidence often cited includes Huawei’s 2023 Mate 60 Pro, which featured a seven nanometer chip produced by Semiconductor Manufacturing International Corporation, the Chinese company known as SMIC, despite the absence of cutting-edge ASML equipment, and more recent developments such as China’s high-performing DeepSeek AI model and Beijing’s 2025 directive instructing major tech firms (including ByteDance and the Alibaba Group) to stop purchasing Nvidia’s RTX Pro 6000D and H20 chips. The government has cited data security concerns and urged firms to justify any purchase from Nvidia when domestic alternatives are available.

There is growing evidence of Chinese firms acquiring high-end AI chips indirectly via third countries and engaging in what is called chip smuggling: The more sprawling export controls are, the harder they are to enforce and the greater the incentives to circumvent them. Even the most effective export controls can only “throw sand in the gears” of China’s technological efforts. However, some commentators argue that China’s technological strength has emerged through lax US enforcement of export controls and that the solution is to give enforcers more resources and to more consistently and thoroughly execute existing rules. In particular, BIS largely relies on one-off approvals of exports to particular customers and reactively responds to violations of export controls after the fact, with little capacity to anticipate them. However, the AI Action Plan proposes to boost enforcement of export controls by ensuring better collaboration between US agencies to monitor where AI chips end up and to detect circumvention. However, BIS efforts to police export controls effectively will become even more difficult if the affiliate rule comes into effect, which will significantly increase the number of companies subject to controls.

Given these dilemmas and uncertainty about the current weighting of US policy objectives, assessing the efficacy of export controls remains difficult. It is impossible to know how China’s technological capabilities would have evolved absent US restrictions, or whether the motivation they provided to accelerate domestic innovation outweighed the obstacles they created.

Impacts of US AI export controls on Europe

The EU and the United States rely on each other for access to AI chips. The latter has many of the world’s most sophisticated chip design firms (e.g., Amazon, Apple, Qualcomm, Google) and providers of chip design software (e.g., Synopsys and Cadence). The EU, on the other hand, is home to ASML, which has a significant lead over firms from elsewhere in the world in producing lithography machines which print circuits onto chips. IMEC, one of the world’s leading semiconductor research and development hubs, based in Belgium, plays a vital role in advancing the AI stack through its work in chip research and manufacturing. US companies benefit from access to European cutting-edge semiconductor innovations, collaborative R&D, secure supply chains and accelerated commercialization. Other components of high-end chipmaking such as specialized chemicals and the lenses used in lithography supply chains are also provided by European firms. Europe’s challenge is therefore not the absence of leverage, but the difficulty of mobilizing it collectively in support of a coherent strategy.

While the EU currently has fewer sectors requiring the highest-end chips compared with the United States, the bloc has clear ambitions to catch up with the United States and China in the AI race. The EU’s AI Continent Action Plan recently emphasized the importance of AI factories—sites equipped with supercomputers and data centers, where developers and start-ups can develop and test AI models. As Europe currently accounts for just 4 percent of the world’s AI computing power, these factories are designed to rival the platforms offered by major non-European tech companies by providing access to computing power, data, and talent, particularly for AI start-ups and small and medium enterprises. Europe also has prioritized the semiconductor industry as crucial to its economic competitiveness and has set an ambitious target of achieving a 20 percent share of the global market by 2030. Meeting this goal will require significant investment and stable, predictable access to foreign markets. The EU therefore plans to supplement its initial Chips Act (which provided subsidies for various parts of the EU semiconductor sector, with at best modest success) with an upcoming Chips Act 2.0. At the same time, the EU’s economic security strategy calls for the strengthening of the continent’s export controls, investment security, and research security measures. Europe has also begun regulating AI use: Despite criticism from the US government and industry, the European Commission maintains that the AI Act will be enforced, though it proposes to delay implementation of most of the law’s rules on AI deployment by one year and provide additional guidance to mitigate impacts on innovation. More broadly, European AI policy will continue to be shaped by commitments designed to avoid higher US tariffs, including agreements to purchase large volumes of US AI chips and to invest hundreds of billions of euros in strategic US sectors. In this context, influence over AI governance will depend less on regulatory approaches alone and more on control over key “choke points” in the AI value chain—including advanced lithography, chip design, compute infrastructure, and critical software ecosystems.

In terms of past export control collaboration, European countries and the United States worked well together to impose export controls on Russia. Since the invasion of Ukraine in 2022, the EU and the United States coordinated their export controls to cut Russia off from dual-use and high-tech goods. Together, they created a Common High-Priority List of restricted items (e.g., computer numerical control machines and microelectronics) and aligned entity sanctions to disrupt Russia’s military-industrial procurement. Export-control coordination has been more difficult when it comes to China, however. In the Biden administration, the US government had seemed to take a tougher approach on export controls than Europe vis-à-vis China’s advances in critical technologies, reportedly proposed to apply the Foreign Direct Product Rule (a provision of US law allowing Washington to control exports of foreign-produced items made using US technology or software) to the Netherlands—home of ASML—unless it imposed tighter restrictions on exports to China.

There had been considerable speculation among some European policymakers that previous export controls were in fact a form of US industrial policy to favor US producers of chip-making tools. While the Dutch government applied stringent controls on ASML, some US chip-making equipment exporters were granted export licences on a regular basis. US policymakers could point, however, to how the proportion of revenue ASML earned from China had been steadily increasing. US pressure on the Netherlands extended beyond ASML. The Nexperia case—in which the Dutch government, following pressure from US officials, ordered the Chinese-owned semiconductor firm to divest from its Dutch chip-making plant on national-security grounds—provided another example of Washington steering allies to take a tougher approach on China.

Although the administration changed, these concerns about US influence over Europe’s technology and AI sectors have persisted. US export controls do not just impact American companies but are also drawn widely enough to restrict how third countries and blocs like the EU do business. This is understandable to some degree; to become more effective, US export controls inevitably have to expand to entangle US allies in order to avoid circumvention risks, and the second Trump administration quickly highlighted its desire to “identify and eliminate loopholes in existing export controls.” In particular, the US has been concerned that some European countries do not have equivalents to US export control tools, such as the Foreign Direct Product Rule. For its part, Europe has been subject to US export controls, which has put it in a weak negotiating position regarding US demands for compliance with unilateral export controls. Europe is particularly vulnerable to US export controls due to their extraterritorial reach, which can impact European companies that use US-origin technology, software, or components. Such firms face compliance risks, supply chain disruption, and potential fines.

The excessive use of such legal tools poses broader geopolitical risks to the United States, however, particularly if the use of those tools is not coordinated with allies. For example, Europeans fear that the Trump administration may use export controls as bargaining chips in bilateral negotiations with China, relaxing or tightening them depending on the outcome. This would mean that the European position would depend on the outcome of US political negotiations in which they are not involved and which might not reflect their interests. Concerns that the United States is an unreliable partner, particularly regarding its tendency to make policy decisions affecting Europe’s access to US technology, is part of the growing European desire to achieve “digital sovereignty” and reduce reliance on the US AI stack to reduce such unpredictability.

Complicating the EU’s efforts to boost its sovereignty and ability to withstand US export controls is Europe’s objective of boosting its economic competitiveness. This is likely to require more diffusion of global technologies in the short to medium term, to boost European firms’ productivity, and therefore have some tension with efforts for the EU to become more technologically self-sufficient. Many economic experts and some successful European tech firms argue that the EU should focus less on self-sufficiency and more on “building on top” of global technologies like AI in a bid to control the technological frontier in at least a few sectors and use this leadership as geopolitical leverage.

More generally, the Trump administration’s AI Action Plan calls for exporting the full “American AI stack,” which includes US chips, AI models, and software. This appears intended to ensure that countries acquiring parts of this AI tech stack would increase their dependencies on the United States, which would pose a significant challenge to Europe’s attempt to achieve more digital sovereignty. These plans are likely to require the EU to continue to acquire some US inputs, while increasing its control over other parts of the value chain. For example, the European Commission is providing support for up to five AI Gigafactories as parts of its sovereignty agenda—even though such gigafactories are likely to rely heavily on US-designed AI chips.

The EU has developed tools that could constrain the export of high-end AI chips to China, particularly given the inactivity of the Wassenaar Arrangement (an export control agreement on dual-use goods that includes Russia, which has been unable to make decisions under this accord since that country’s full-scale invasion of Ukraine). These tools include strengthened security rules around dual-use goods, as well as new or strengthened investment-screening and export-control mechanisms at the member-state level. For example, several EU member states, including the Netherlands, Spain, Germany, and Lithuania, have adopted national control lists under the EU Dual Use Regulation. These lists target emerging dual-use technologies such as semiconductors, quantum computing, and additive manufacturing. The European Commission released a white paper on export controls in January 2024 and a new Economic Security Strategy in December 2025, which set out various reforms that could be implemented to strengthen and better coordinate European export control policies. In a thinly veiled reference to US–China tensions, the European Commission justified the need for a common export control policy by stating that “the lack of a common EU voice exposes individual Member States to geopolitical pressures.” Despite many actors, including several business stakeholders, being willing to implement reforms, the ability of Europeans to create a common export control system is hindered by the fact that it touches on security matters that fall within the remit of member states. Some member states have already stated that they prefer ad hoc formats, declaring that “the Commission has no role to play” in export controls. Furthermore, at the EU level, the design of tools like the Anti-Coercion Instrument make it difficult for the EU to act like a true geopolitical actor even when this means taking a more assertive approach than the least-willing member states.

EU-US cooperation and export control policy design

For the EU, several difficulties hinder cooperation with the United States on export control policy. The first is the unpredictability of US policy on China. There seems to be disagreement within the US administration about how to treat China, as reflected in variations in the direction of export control policy at times as well as how policy trade-offs are managed differently. European leaders are concerned that any aspect of US policy could change in any US-China deal. The second difficulty is that the United States and the EU do not (yet) have a similar set of policy objectives when it comes to China. Overall, Washington is far more willing to openly treat China as a threat; in Brussels, the position on China is evolving, but different member states still have different positions, making a clear and unified European position difficult to coalesce. A third challenge is that US policy, at least to some degree, seems to give little weight to the interests of US allies, seeing them in transactional terms, as “export destinations,” rather than as partners who have their own interests to protection and who should share in the benefits of cooperation.

Over time, as AI capabilities increasingly diffuse beyond hardware into models and software, export controls alone are likely to become less effective. This will increase the importance of transatlantic cooperation on standards, monitoring, and downstream use of AI systems.

If constraints on cooperation can be overcome, then the question is what form cooperation could take. Under the Biden administration, the EU and the United States had formal forums for cooperation on AI security including in the Trade and Technology Council (TTC) working group on AI, and in a Joint Roadmap for Trustworthy AI and Risk Management. However, in the Trump administration, the TTC appears unlikely to return in its previous form, and it is unclear whether the US leadership is willing to create different formats to align transatlantic preferences on AI governance and risk management. Pax Silica appears to be an explicitly US-led initiative, which the United States wishes European countries to join, rather than a genuine forum for EU-US dialogue. At the political level, achieving a grand-strategic agreement on AI at the transatlantic level appears unfeasible. A more pragmatic approach would be to consider what steps Washington and Brussels could take to align their export controls and, potentially, converge their respective economic security agendas.

The US AI Action Plan and the executive order to promote the export of American AI technology call for robust enforcement of export controls. This reinforces the administration’s intention to oversee and ensure compliance with AI exports internationally, rather than simply pursuing a laissez-faire and deregulatory approach. However, while the Trump administration’s AI Action Plan seeks to encourage deals with allies to coordinate technology controls, it also signals a move toward more assertively ensuring allies align with the United States—including through trade tools to discourage countries from undermining the objectives of US export controls and the potential use of secondary sanctions to enforce compliance with key AI export restrictions. A question for Europe is how it can realistically ensure its interests are taken into account under this approach. Since the current US administration is unlikely to bend to European interests merely out of goodwill, Europe must more effectively develop technological leadership and “choke points” in AI supply chains to give itself leverage (and, more generally, for Europe to avoid dependencies on the United States in even more strategic areas like defense, which greatly impact the EU’s willingness to assert its own interests). The United States may in the future require closer EU-US cooperation to achieve its policy objectives, and so it is in the US interest to lay the groundwork for future collaboration in cases where that makes sense. An inability for the United States and Europe to effectively coordinate regarding export controls could have a particularly damaging impact in the context of US-China technological competition. The US desire to export its AI and technological infrastructure to third countries makes robust enforcement mechanisms even more necessary. Deploying US AI technology abroad carries significant leakage risks, including the diversion of advanced chips, exposure of model weights, and unauthorized access through intermediaries. Therefore, agreements with allies are essential to reduce monitoring, enforcement, and audit workloads. The EU should recognize, in this context, concerns from the United States that not all of its member states take the same robust approach or willingness to share the workload of effective monitoring and enforcement. This will become even more important because enforcing export controls, which has historically been challenging, is likely to be even more complex in the AI era. It is difficult to prevent foreign governments or companies from developing AI models using computing infrastructure or software advancements made by open-source groups.

Furthermore, it would be incomplete to view export controls in isolation from the broader economic security agenda. Export controls are protective tools that work best when coordinated rather than issued unilaterally. This is particularly true today, given the complexity of the AI stack and the continuous evolution of AI technology and the market, which are spread across multiple jurisdictions. Export controls are most effective when used alongside other protective measures, such as investment-screening mechanisms, to close loopholes and improve the enforcement of protective measures. Moreover, export controls do not guarantee, by themselves, a technological advantage; indeed, they can sometimes have a negative impact on the most advanced industries themselves, and so must be explicitly linked to an agenda promoting innovation and technological competitiveness.

The transatlantic relationship faces both opportunities and challenges in this regard. The EU and the United States would be well advised to collaborate, ensuring complementarity and respecting each other’s competitive advantages, to achieve a shared technological advantage. In the long run a club approach—comprising the EU, United States, and potentially other like-minded democracies with a strong presence in AI value chains such as Taiwan, Japan, and South Korea—could collaborate around building a common market, accepting their respective comparative advantages, avoiding unilateral bargaining tools, sharing the burdens of economic and national security policies, and adopting protective measures to avoid harmful dependencies.

However, such an approach is not realistic in the short term. There is no doubt that the transatlantic relationship is asymmetrical, at least in the AI sector (and probably in other strategic sectors), with the risk that the United States could exploit this to its advantage while European countries could find themselves derisking with respect to the US AI stack to minimize dependencies. To mitigate the risk of transatlantic fragmentation in AI, it is crucial to continue reflecting on and discussing how to align economic security agendas at the transatlantic level, striking a balance between strategic protection and promoting innovation. In the short run, cooperation could be more modest and pragmatic. It could rely on regular (structured or informal) policy dialogue to ensure each side understands the other’s approach and to avoid surprises. The two sides could also undertake scenario planning and risk assessments to identify AI security risks, potential policy tools to mitigate them, and to predict and prepare responses to potential retaliation. They could also boost public-private engagement to ensure policies are informed by real-world capabilities and risks. Most importantly, the EU and United States should identify and work on areas of cooperation such as adopting joint measures to stop China from swamping global markets for legacy chips, driving out non-Chinese suppliers, and creating new dependencies. Both the EU and United States have manufacturers of legacy chips and customers which need these legacy chips. This could serve as an area to help demonstrate the benefits of a joint approach and could help convince the US administration that the EU can be a partner, not just a useful export destination, in managing national security risks in AI. One practical step could be the creation of a small, informal EU-US coordination group on AI export controls, focused on information sharing, risk monitoring (e.g., chip diversion), and aligning enforcement approaches in specific sectors.

Antonio Calcara is the head of the Geopolitics and Technology Programme at the Centre for Security, Diplomacy and Strategy and a research professor at Vrije University Brussels, where he is leading the European Research Council project “Competition in the Digital Era (CODE): Geopolitics and Technology in the 21st Century.“

This issue brief benefits from the insights of discussants at an online roundtable on EU-US regulatory co-operation hosted jointly by CERRE and the . However, the contents of this brief are attributable only to the author.

Providing high-quality studies and dissemination activities, the Centre on Regulation in Europe (CERRE) is a not-for-profit think tank. It promotes robust and consistent regulation in Europe’s network, digital industry, and service sectors. CERRE’s members are regulatory authorities and companies operating in these sectors, as well as universities.

CERRE’s added value is based on

• its original, multidisciplinary, and cross-sector approach covering a variety of markets (e.g., energy, mobility, sustainability, technology, media, and telecommunications);
• the widely acknowledged academic credentials and policy experience of its research team and associated staff members;
• its scientific independence and impartiality; and
• the direct relevance and timeliness of its contributions to the policy and regulatory development process impacting network industry players and the markets for their goods and services.

CERRE’s activities include contributions to the development of norms, standards, and policy recommendations related to the regulation of service providers, to the specification of market rules, and to improvements in the management of infrastructure in a changing political, economic, technological, and social environment. CERRE’s work also aims to clarify the respective roles of market operators, governments, and regulatory authorities, as well as contribute to the enhancement of those organizations’ expertise in addressing regulatory issues of relevance to their activities.

The promotes constructive leadership and engagement in international affairs based on the Atlantic community’s central role in meeting global challenges. The council provides an essential forum for navigating the dramatic economic and political changes defining the twenty-first century by informing and galvanizing its uniquely influential network of global leaders. The —through the papers it publishes, the ideas it generates, the future leaders it develops, and the communities it builds—shapes policy choices and strategies to create a more free, secure, and prosperous world.

The ’s Europe Center conducts research and uses real-time analysis to inform the actions and strategies of key transatlantic decision-makers in the face of great-power competition and a geopolitical rewiring of Europe. The center convenes US and European leaders to promote dialogue and make the case for the US-EU partnership as a key asset for the United States and Europe alike. The center’s Transatlantic Digital Marketplace Initiative seeks to foster greater US-EU understanding and collaboration on digital policy matters and makes recommendations for building cooperation and ameliorating differences in this fast-growing area of the transatlantic economy.

Image: A computer chip with illuminated connections. Courtesy of iStock