Artificial intelligence is no longer a domain governed by tech optimism and unregulated diffusion. Since 2023, the international community has entered a legal epoch in which the export of AI technologies is regulated as a matter of national security, economic strategy and geopolitical contestation. The United States, the European Union and China have each constructed regulatory frameworks that reflect their values, priorities and strategic ambitions. These regimes do not merely seek to control the movement of technology across borders, they embody competing visions of technological governance. Understanding these divergent legal architectures is essential for policymakers academics and corporate strategists alike. This article offers an expert comparative analysis of AI export regulation as a distinct legal regime, revealing the conceptual underpinnings, operational mechanisms and global implications of each model.
To appreciate the significance of AI export regulation one must first recognise the fundamental shift in how states perceive AI. Where once artificial intelligence was primarily an economic asset, it is now understood to be both a core enabler of military capability and a strategic determinant of national power. AI underpins everything from autonomous systems and intelligence analysis to critical infrastructure resilience and economic competitiveness. Viewed through this lens, export regulation ceases to be a narrow trade issue and becomes a component of national security architecture.
The US Framework
In the United States the regulation of AI exports has been integrated into the broader architecture of national security law and the export control regime administered by the Department of Commerce’s Bureau of Industry and Security. Historically, export controls focused on tangible items like semiconductor chips, advanced materials and defence articles. By 2025, however, the regulatory scope had expanded to include software and model weights of AI systems, a recognition that the true strategic value of AI resides in its intangible computational parameters as much as in physical hardware.
The legal foundation for these controls is the Export Administration Regulations, implemented under the authority of the Export Control Reform Act. The United States introduced new classification categories that capture “closed-weight” AI models trained above defined computational thresholds. These classifications are not arbitrary. They are grounded in an assessment of dual use risk, reflecting the degree to which certain AI models could enable capabilities with clear national security implications. A model capable of high-complexity pattern recognition or autonomous decision support, for instance, poses a different risk profile than a basic classification algorithm.
What makes the US framework distinctive is not merely its inclusion of software but the legal logic it applies to jurisdiction. The United States asserts control not only over exports originating within its own territory, but also over foreign produced models that are direct products of US controlled hardware or software tools. This extension of jurisdiction demonstrates a willingness to interpret export control statutes expansively, equating economic leverage with legal reach.
Equally significant is the licensing regime. Exports to allied jurisdictions may benefit from tailored exceptions provided that the end user and end use satisfy stringent security criteria. Conversely, exports to high-risk destinations face a presumption of denial. This binary structure is designed to reinforce geopolitical alignments through legal instruments.
The EU Framework: Regulatory Equivalence and Democratic Norms
In contrast to the US model’s emphasis on national security thresholds, the European Union’s approach to AI export regulation is embedded within its broader commitment to normative regulation and risk governance. The EU’s legal framework does not rely primarily on numerical model weights or computational thresholds. Instead it is concerned with categorising AI capabilities according to risk profiles and aligning export controls with core European values such as fundamental rights protection, human dignity and democratic accountability.
The EU framework is best understood in relation to the Artificial Intelligence Act, which classifies AI systems into unacceptable, high, limited and minimal risk categories. While the AI Act governs domestic deployment, its influence extends to export regulation through coordinated mechanisms that seek to ensure that exported technology adheres to European ethical standards.
In practice this means that high risk AI systems subject to stringent oversight within the EU cannot be exported without demonstrating compliance with the same standards abroad. This regulatory equivalence model reflects a legal commitment to normative consistency: technology that cannot operate within the EU without safeguards should not be disseminated externally without equivalent safeguards.
The legal challenge for the EU lies in translating domestic regulatory values into extraterritorial effect without overreaching international norms. Unlike the United States the EU cannot rely on a single export control statute with broad national security exceptions. Its authority is derived from a combination of trade law, internal market harmonisation powers and international commitments. Consequently the EU often uses export licensing as a complementary tool, activated when AI systems intersect with existing lists of dual use technologies or when aligned with sanctions policy rather than as a standalone regime.
The Chinese Framework
China’s approach to AI export regulation is deeply shaped by its philosophy of cyber sovereignty and state centred economic planning. The legal architecture governing AI exports in China is not yet codified in a single all encompassing statute but is emerging through a constellation of laws relating to data security, cybersecurity, export controls and national intelligence.
The core principle of the Chinese model is the protection of information assets as an extension of state security. The Cybersecurity Law and the Data Security Law impose strict controls on the transfer of data and algorithmic models that are designated as important or critical information infrastructure. AI models trained on data deemed sensitive or strategic fall within this category. Export control guidance emanating from the Ministry of Commerce emphasises the prevention of technologically induced risks and the safeguarding of state interests.
From a legal perspective, China’s export regulations justify control on the basis of state necessity and industrial safeguarding rather than an external threat as in the US case. The rhetoric of threats is internal: uncontrolled diffusion of AI technology could undermine domestic stability, compromise national data assets and dilute China’s strategic advantage. Consequently Chinese firms seeking to export advanced AI capabilities must obtain government approval if the technology intersects with categories enumerated in export control guidance, notably those involving encryption autonomy algorithms or deep learning models trained on large scale datasets.
The Chinese model is also characterised by a preference for bilateral technology agreements. Where possible Beijing negotiates with partner states to establish reciprocal arrangements that grant market access while ensuring that Chinese regulatory concerns are respected. This reflects a legal culture that views international agreements not as constraints on state power but as mechanisms to institutionalise state control.
Comparative Legal Analysis: Jurisdiction, Sovereignty and Compliance
Comparing these three frameworks reveals contrasting legal logics. The United States approach is rooted in a national security priority that foregrounds control over capability diffusion. It accepts expansive jurisdictional claims where strategic risk is identified. The EU approach is rooted in normative regulatory coherence, concerned with exporting values as much as technology and managing risk through equivalence. China’s approach is rooted in sovereign security and industrial statecraft, prioritising domestic control and bilateral governance over multilateral regulatory harmonisation.
These differences have practical consequences for global technology governance. A firm in Singapore seeking to export a high performance AI model must navigate US control lists if the technology incorporates US origin components, EU obligations if the software is developed or marketed in the EU and Chinese export controls if the model was trained on data subject to Chinese security provisions. The legal complexity grows when these frameworks interact with third country laws, digital trade commitments and emerging cooperative mechanisms such as the proposed OECD framework for AI principles.
One of the most fascinating aspects of this comparative landscape is how each regime views the object of regulation. The United States focuses on capability thresholds and potential misuse scenarios, the EU on ethical and human rights outcomes, and China on state defined categories of information assets and strategic technology classes. These differing ontologies shape export control practices in subtle ways. For example a model that the EU considers high risk because of its potential impact on fundamental rights might not trigger US controls if it falls below a computational threshold. Conversely a model that is tightly regulated in the United States for national security reasons might be exportable from China with minimal restriction if it does not meet China’s domestic definition of strategic asset.
Legal Compliance and the Future of AI Export Governance
The fragmentation of AI export regulation poses challenges for international commerce and legal compliance. Companies must build multi jurisdictional compliance functions that integrate export control screening, risk assessment aligned with ethical governance and strategic licensing pathways. This is not a simple checklist exercise but a dynamic legal intelligence function that tracks evolving thresholds, licensing criteria and enforcement patterns.
For states, the challenge is equally profound. If the global AI ecosystem is governed by divergent legal regimes with competing priorities, there is a risk that innovation will be stifled in some jurisdictions while proliferating unchecked in others. This raises questions about regulatory convergence and possible frameworks for interoperability. Possibilities include negotiated treaty instruments focusing on high risk AI systems, joint licensing pilots among trusted partners or harmonised transparency standards that reduce the complexity of compliance while preserving core policy objectives.
The comparative legal frameworks of the United States, European Union and China also illuminate deeper questions about the future of international law. Export regulation once centred on physical goods and services is now grappling with intangible technologies that embody knowledge, data and adaptive behaviour. This transition from goods to algorithms requires legal imagination and doctrinal evolution. Traditional concepts of jurisdiction, control and enforcement must adapt to a world in which software is both the object of regulation and the medium of enforcement.
In this sense AI export regulation is not merely an extension of export controls. It represents a new legal regime that synthesises trade law, national security law, human rights law, digital governance and industrial policy. Its development will shape technological trajectories and geopolitical alignments for decades. The United States, the European Union and China have each staked out distinct legal paths. Whether these paths converge through practice or clash in courts, councils and capitals will be one of the defining legal narratives of the twenty first century.
For advocates and scholars the imperative is to engage with this regime not as an abstraction but as the living architecture of law in an era defined by data intelligence and strategic competition. If artificial intelligence transforms societies, the legal frameworks that govern its export will shape the contours of global power in ways that are only now beginning to be understood.
Export Controls vs WTO Law Are National Controls Eroding the Multilateral Trade System
In 2026 the tension between unilateral export controls and the multilateral trading system governed by the World Trade Organization is no longer an abstract academic concern. What once seemed a niche area of trade law has erupted into a fundamental challenge to the architecture of global commerce. Nations are increasingly deploying export controls not merely to regulate the movement of dual use goods, but as instruments of strategic statecraft. Advanced semiconductors artificial intelligence and critical technologies have become geopolitical fulcrums. In response, states from Washington to Beijing have asserted export control regimes that operate with considerable extraterritorial effect. The question that arises with urgency is whether this trend undermines the rule based order embodied in the WTO, and if so what that means for the future of global trade governance.
At the heart of the matter lies a tension between sovereign prerogatives and institutional commitments. Export controls are an expression of sovereign authority, rooted historically in national security concerns. WTO law is an expression of a collective commitment to non discriminatory trade and mutual legal obligations among member states. The relationship between these two legal orders is neither simple nor static. It is a dialectic shaped by political imperatives, economic power and evolving conceptions of security in an interconnected world.
To appreciate the emerging friction one must first understand the legal foundations of both regimes. Export controls have historically been enforced under domestic statutes such as the United States Export Control Reform Act, the European Union Dual Use Regulation and similar legal instruments in other jurisdictions. These frameworks empower states to restrict the cross border movement of goods, software or technology deemed sensitive for national security, foreign policy or public order objectives. In essence export controls are regulatory police powers applied at the frontier of sovereign jurisdictions.
By contrast World Trade Organization law is a treaty based legal order built on principles of non discrimination, market access and transparency. Obligations under the General Agreement on Tariffs and Trade include most favoured nation treatment and national treatment, which require equal treatment of like products irrespective of origin or destination. WTO disciplines on quantitative restrictions and general elimination of tariffs are mechanisms to prevent arbitrary trade distortion.
The legal premises appear to pull in different directions. Export controls deliberately restrict trade in response to risk. WTO law facilitates trade in response to the benefits of market openness. The WTO recognises exceptions, most prominently in Article XXI of the GATT, which allows members to adopt measures they consider necessary for the protection of essential security interests. But this exception is not a free pass. It has been rarely litigated, partly because adjudicating security claims implicates sensitive political judgements and partly because there is no clear jurisprudential standard on how such exceptions should be interpreted.
In recent years the invocation of national security exceptions has become more frequent and more contentious. The United States, in particular, has integrated export controls into its broader strategy to contain the rise of peer competitors in technology sectors such as semiconductors artificial intelligence and quantum computing. The Commerce Department’s expansion of export control lists to cover high computing technology and AI model weights is a striking illustration. These are not classical munitions. They are economic inputs whose export restrictions aim to prevent strategic adversaries from acquiring competitive capabilities. The extraterritorial reach claimed by the United States in such rules, especially through the Foreign Direct Product Rule, has generated anxiety among trade partners about the scope of US legal authority and its impact on non US entities.
Europe’s approach has been more measured but no less significant. The European Union Dual Use Regulation has been amended to capture emerging technologies with potential military application. Divergence between EU and US lists, and the absence of consistent multilateral coordination, has created complexity for exporting firms and uncertainty for global supply chains. China’s own export control regime, articulated through its Export Control Law, is couched in the language of protecting national security and economic interests and grants Beijing broad discretion to decide what technologies require authorisation before export.
These regimes operate in parallel with WTO commitments, but the relationship is uneasy. Export controls often have the effect of restricting trade in ways that appear inconsistent with WTO obligations. For example when a state restricts the export of a high technology component to a particular foreign destination, the effect may be functionally equivalent to a quantitative restriction on exports, which WTO law generally prohibits.
Defenders of export controls argue that security imperatives must take precedence and that WTO law, through Article XXI, provides sufficient legal space for these measures. Critics counter that Article XXI has been deployed as a shield rather than a narrowly construed exception. When virtually any technology with potential strategic application is categorised as a security risk the exception swallows the rule. This creates an ambiguity in WTO law that is being exploited as a legal loophole, undermining the predictability and mutual restraint that the WTO was designed to guarantee.
A striking example of this tension is the semiconductor supply chain. Advanced node chips are subject to export controls that vary significantly among major players. The United States restricts exports of semiconductor manufacturing equipment and high performance computing platforms to countries it deems strategically sensitive. China has responded by restricting exports of gallium and germanium, key materials used in chip production. Both sets of controls have significant effects on global trade flows. In theory these measures could be challenged at the WTO as unlawful export restrictions. In practice, the political cost of doing so, coupled with the blurred line between economic competition and national security, has deterred formal dispute settlement.
This dynamic raises a broader question about the functionality of the WTO dispute settlement system in an era of pervasive security justifications. If members can shelter export controls behind security exceptions without meaningful review, WTO law risks becoming hollowed out. The result is legal fragmentation and regulatory regionalisation. Instead of a cohesive multilateral trade order, the world could drift towards competing blocs with overlapping and contradictory rules.
Some scholars have proposed that WTO jurisprudence should develop clearer standards for reviewing security exceptions. Others suggest that trade and security should be disentangled, with new multilateral mechanisms designed to handle technology controls outside the WTO framework. Both approaches recognise that the historical assumptions underpinning the WTO, such as a primary focus on tariffs and quantitative restrictions, do not neatly accommodate the complexities of twenty first century technology trade.
Yet there is also a countervailing force that points towards adaptation rather than erosion. Emerging practices among WTO members show that export controls are being discussed within institutional venues and technical committees. Countries have begun to share classification lists and coordinate on licensing practices with a view to reducing unnecessary trade disruption. These efforts suggest that even in areas of strategic tension there is appetite for some degree of harmonisation.
The legal future of export controls and WTO law may depend on whether members are willing to negotiate new normative frameworks that reconcile legitimate security concerns with the need for a stable trade order. An example might be an agreement on dual use technologies that establishes clear criteria for what constitutes a legitimate security risk, balanced against commitments to transparency, temporal limits on measures and meaningful consultation. Such an agreement would not eliminate export controls, but it would constrain their use in ways that preserve core principles of non discrimination and predictability.
In the absence of such normative evolution, national export controls will continue to push the boundaries of WTO law. Firms engaged in international commerce will face increasing compliance complexity as they navigate a landscape where overlapping controls reflect not just economic policy but geopolitical rivalry. Legal advisors will need to become adept not only at understanding domestic export statutes but at mapping their interaction with multilateral obligations and the evolving norms of international economic governance.
The central paradox is that export controls, intended as instruments of national security, may in aggregate erode the very legal infrastructure that underpins global economic stability. A fragmented regime in which states unilaterally assert broad export controls without effective multilateral constraints could yield short term strategic gains but long term systemic instability.
Whether the multilateral trading system can adapt remains an open question. What is certain is that the debate over export controls and WTO law will define the contours of international economic law for years to come. The stakes are not merely legal neatness or trade flows. They are about how an interconnected world chooses to balance sovereign security, economic interdependence and the rule of law itself.