The international community gathered at the AI Impact Summit 2026 in New Delhi, India, on 19 February 2026, reconfirmed that the future of artificial intelligence (AI) must be one of inclusion and safety. On the economic front, the Summit Declaration signatories noted that the widespread adoption of AI has unique potential for economic and social development. The aim is to scale up and replicate best practices globally, facilitating a digital ecosystem that is accessible and adaptable to all sectors [‎‎12]. 

The Declaration was signed by 92 states and international organizations, including the EU, sending a clear signal that AI has a major impact on the global economy and well-being.

During the Summit, the EU AI Act, the world’s most comprehensive horizontal AI regulation, was harshly criticized. Particularly, it was called overregulating, anti-entrepreneurial, and anti-innovational. 

Despite this, the EUs goal is improving the functioning of the internal market and promote the uptake of human-centric and trustworthy AI, while ensuring a high level of protection of health, safety, fundamental rights enshrined in the Charter of Fundamental Rights of the EU, including democracy, the rule of law and environmental protection, against the harmful effects of AI systems in the EU and supporting innovation [‎17, Article 1(1)].

EU AI Act establishes a risk-based approach that implies stricter rules in case of the higher risks of causing damages, and softer rules in the eventuality of lower risks or their absence, respectively. The Act defines the following levels of risk for AI systems: 1) unacceptable risk (prohibited AI systems); 2) high risk; 3) limited risk; and 4) minimal risk [‎7].

Unacceptable risk to fundamental rights and the EU values exists in the case of the following AI-practices listed in Article 5 of the EU AI Act, that shall be prohibited: harmful manipulation and deception; harmful exploitation of vulnerabilities; social scoring; individual criminal offence risk assessment and prediction; untargeted scraping to develop facial recognition databases; emotion recognition; biometric categorization; biometric categorization; and real-time remote biometric identification [‎17, Article 5(1)]. 

As for high-risk AI systems, to be qualified as such, they shall meet cumulatively two conditions: (a) the AI system is intended to be used as a safety component of a product, or the AI system is itself a product, covered by the EU harmonization legislation listed in Annex I to the EU AI Act; (b) the product whose safety component pursuant to point (a) is the AI system, or the AI system itself as a product, is required to undergo a third-party conformity assessment, with a view to the placing on the market or the putting into service of that product pursuant to the Union harmonization legislation listed in Annex I [‎17, Article 6(1)].

The EU AI Act then outlines the areas in which high-risk AI systems are listed: biometrics, critical infrastructure, education and vocational training, employment, essential services, law enforcement, migration and border control, administration of justice and democratic processes, and safety components subject to EU harmonization legislation that are required to go through a third-party conformity assessment under the EU harmonization legislation [‎17, Annex III]. 

Other AI systems are considered to have limited risk, for instance, those interacting with consumers and GPAI. Providers and deployers of respective AI systems bear transparency obligations. For instance, providers shall ensure that, while interacting with AI systems, natural persons shall be aware of and sure about the fact they interact with an AI system. Also, providers of AI systems, including GPAI ones, shall guarantee that AI-generated content is identifiable [‎17, Article 50(1)-(2)]. 

Regarding deployers of AI systems, they shall inform natural persons that their personal data are properly processed, and shall disclose AI-generated and manipulated content.

Finally, the EU AI Act does not provide rules neither for AI deemed having minimal or no risk at all, nor defines AI systems falling into this risk level.

As for trade impact, the European Commission highlighted that the risk-based regulatory approach does not create unnecessary restrictions to trade, whereby legal intervention is tailored to those concrete situations where there is a justified cause for concern or where such concern can reasonably be anticipated in the near future [‎8].

The EU AI Act must be seen not just as a solid AI regulatory landscape but also as an instrument affecting international trade due to its extraterritorial effect. It does not distinguish between foreign and domestic providers, as long as they place on the market or put into service AI systems, among those and general-purpose AI models in the EU [‎17, Article 2(1)(a)]. Also, AI systems deployers, whose place of establishment or they are located in a third country, are subject to EU regulation. Authorized representatives of providers, not established in the EU, are also subjects of the EU AI Act. In case AI systems affect persons located in the EU, external providers would also become the subject of regulation, since the norm does not specify the origin of providers envisaged [‎17, Article 2(1)].

The degree of affection of international trade by the EU AI Act shall be made through the prism of the World Trade Organization (WTO) main treaties: General Agreement on Tariffs and Trade (GATT), General Agreement on Trade in Services (GATS), and the Agreement on Technical Barriers to Trade (TBT Agreement).

What is AI in the international trade context: good or service?

In assessing the compatibility with WTO disciplines, the first question to address is whether international trade involving AI can be categorized as trade in goods or in services [‎18, p. 16].

The WTOs moratorium on customs duties for electronic transmissions (moratorium), first adopted in 1998 and repeatedly renewed [‎28], has preserved tariff-free digital trade but also deferred the categorization of the AI dilemma. By postponing that determination, the moratorium creates constant legal uncertainty and asymmetric effects on developing and developed countries and increases the risk of regulatory fragmentation if WTO Members adopt unilateral measures in the absence of a clear multilateral rule.

At the 13th Ministerial Conference (MC), the WTO Members agreed that the moratorium, as well as the Work Programme on Electronic Commerce, would expire at the 14th Ministerial Conference or on 31 March 2026, whichever comes first [‎28].

WTO Members concluded at the MC14 that took place between 26-29 March 2026 at Yaoundé, Cameroon, both moratorium and the E-Commerce Work Programme came to end on 31 March 2026. Although WTO Members strived to maintain the practice of not imposing customs duties on electronic transmissions until 31 December 2030 [‎26, Annex, p. 4, para. 5], they failed to agree on the moratorium extension. So, WTO Members could apply customs duties on electronic transmissions that can include AI-driven products. 

However, 66 WTO Members, including the EU, prepared interim arrangements for the Agreement on Electronic Commerce (E-Commerce Agreement) that should be added to Annex 4 of the WTO Agreement. E-Commerce Agreement will serve as a first set of rules in digital trade, aimed to support digital transformation within and across their respective economies through rules-based frameworks [‎25, p. 1, para. 3].

The E-Commerce Agreement provision of interest is Article 11. It first broadly defines electronic transmissions, and this suggests that AI, whether as an autonomous product or integrated in goods or services, is included in the notion.

Parties to Agreement will agree to not impose customs duties on electronic transmissions between a person of one Party and person of another Party. Concurrently, a Party could impose internal taxes, fees, or other charges on electronic transmissions in a manner not inconsistent with the WTO Agreement. 

In light of the AI classification issue, Article 11.5 becomes a bridge between the static nature of duty-free electronic transmissions and the dynamic nature of AI. The provision binds the Parties to revise Article 11 content in the fifth year after the date of entry into force of the E-Commerce Agreement. This should ensure that the current broad definition of electronic transmission does not become a permanent loophole that prevents the re-categorization of AI as a good or service. 

The E-Commerce Agreement, if adopted, seems to sidestep the traditional classification dilemma by shifting the focus from the nature of AI-driven products to the mode of delivery, effectively treating AI as a duty-free electronic transmission rather than good or service.

Theoretically, AI products could be both goods and services. On the one hand, AI can be the part of goods if it is integrated into tangible products, like robotics. 

Nevertheless, the World Customs Organization Harmonized System (HS) Nomenclature of Goods does not mention AI itself or products powered by its technology. Instead, it takes precedence over a product’s physical function and hardware components compared to its internalized digital capabilities. 

There is no name for the technology, rather the system abstracts AI through identifying the types of hardware that allow it to exist or by what kind of machine it occupies. This trait-based logic pattern is applied when all but the most advanced autonomous systems shall go into Heading 88.06 as unmanned aircraft by virtue of their physical nature being an aircraft capable of flight versus the intelligence that runs it [‎23, Chapter 88, Heading 88.06].

Therefore, AI is not regarded by GATT as a standalone ‘good’. Rather, it is deemed as a useful part of a tangible object. As a result, GATT only applies to AI when it is incorporated into a tangible product that is categorized using the HS Nomenclature of Goods.

On the other hand, AI can be incorporated into education, telecommunication, or transportation systems. Yet AI can also be an autonomous service by providing information, knowledge, and ideas to recipients [‎18, p. 17]. Theoretically, GATS rules can be applicable. 

However, the issue of proper classification of AI-driven services was not fully resolved. The classification of services sectors and sub-sectors should be grounded on the WTO Secretariat Services Sectoral Classification List (W/120) [‎27]. 

Nonetheless, this framework remains tethered to an obsolete foundation. Originally derived from the UNs Provisional Central Product Classification (CPC) in 1991, the W/120 was incorporated into WTO practice upon the entry into force of the GATS in 1995. While it initially provided a comprehensive map of the services landscape, it has failed to keep pace with the rapid technological evolution of the last three decades.

Despite the emergence of sophisticated frameworks like the CPC Version 2.1, released in 2015, which explicitly account for modern digital services, the WTO has yet to formally adopt these updates. While W/120 remains the legal benchmark for commitments, CPC Version 2.1 serves as a vital interpretive tool in technical disputes to determine whether nascent AI services fall within specific sectoral classifications.

However, even the reliance on CPC Version 2.1 itself will soon become precarious. In 2025, the United Nations Statistic Division adopted CPC Version 3.0 that is going to supersede Version 2.1. At the time of drafting this article, it has not yet been published.

Among the objectives of the forthcoming CPC is to align the classification with existing economic and technological reality while maintaining conceptual consistency [‎21, p. 3]. The revision process explicitly addressed AI, although not by creating a standalone category, but by updating the Explanatory Notes of existing subclasses to reflect AIs ubiquitous nature.

Particularly, AI will be explicitly recognized within subclasses 83152 Application software provision and 84392 On-line software. The first subclass covers the hosting and management of applications, including, for instance, provision of software-as-a-service (SaaS) and function-as-a-service (FaaS). The inclusion of AI here is crucial for services where AI is bundled with consulting, customization, and system integration. The latter subclass captures software intended for online execution, excluding at the same time game software, and software downloads [‎20, p. 418, p. 442].

The UNs decision to avoid a dedicated subclass to AI may suggest that the respective services shall be classified under the category appropriate to the services to which they are more akin [‎22].

In eventual disputes, the forthcoming CPC Version 3.0 provides a modern interpretive lens to determine whether an AI-driven service falls under a specific product classification, even if the underlying schedule of commitments of any WTO Member relies on the outdated W/120.

Even if the thesis that AI products are attributable to both goods or services is still questionable, the article will support both viewpoints. It will argue the EU AI Act compliance with GATT, GATS, and the TBT Agreement.

EU AI Act and its compatibility with WTO treaties

The EU AI Act could conflict with GATT. In situations when AI-powered goods would be placed on the market, put into service or used in the EU, the risk-based approach would restrict the abovementioned actions. It concerns prohibited AI practices under Article 5 or high-risk AI systems under Article 6 of the EU AI Act. Obviously, hypothetical limitations would be deemed quantitative restrictions under Article XI:1 GATT.

So, the EU should seek justification of trade restrictive measures under Article XX GATT.

The two-tier analysis under Article XX GATS is required for the EU to properly justify eventual measures affecting trade in AI-driven goods. Respectively, the WTO Panel should first determine whether there is any provisional justification for the inconsistent measure under paragraphs (a)-(j). Then, in case the measure is covered by Article XX, the Panel should find whether it meets the requirements of the Article XX chapeau [‎5, para. 123].

In respect of the first step, one of its elements is determining the necessity of the measure at stake. The measures necessity analysis consists in weighing and balancing the following, but not limited to, factors: (i) importance of the objective pursued, (ii) contribution of the measure to the achievement of the objective, and (iii) restrictive impact on international commerce [‎15, para. 7.104]. Also, a comparison between the challenged measure and possible alternatives should then be undertaken [‎5, para. 307].

Accordingly, the EU could invoke the necessity of protection of public morals, public order or human, animal or plant life or health. This arises from the EU AI Act subject matter enshrined in Article 1(1) cited before.

Besides, the EU could justify the measure if it is necessary to secure compliance with GATT-consistent laws or regulations relating to customs enforcement, the enforcement of monopolies operated under paragraph 4 of Article II and Article XVII, the protection of patents, trademarks and copyrights, and the prevention of deceptive practices.

The next step of the two-tier analysis is satisfaction of GATT Article XX chapeau. Specifically, the justification of a measure under chapeau is viable if there is no arbitrary or unjustifiable discrimination between countries where same conditions prevail or disguised restriction on international trade [‎4, para. 118].

Measures adopted under EU AI Act could be also justified under security exceptions established in GATT Article XXI. 

Worth mentioning that measures taken under the EU AI Act could not be justified in peacetime, if the EU would invoke GATT Article XXI (b)(iii) in time of war or other emergency in international relations. This inevitably leads to the question of interpretation of the phrase emergency in international relations.

The phrase was construed in the WTO Panels recent report in light of GATS Article XIV bis (b)(iii). The Panel found that emergency in international relations refers to a state of affairs, of the utmost gravity, in effect a situation representing a breakdown or near-breakdown in the relations between states or other participants in international relations. The Panel is enjoined from assessing the degree of deterioration of international relations, regardless of what caused that deterioration. The emphasis should be put on the gravity of impact of that situation on states or other participants in international relations [‎16, para. 7.306-7.308].

The EU AI Acts interference with GATS is also possible. It may clash with GATS Article XVI:2(a), since some AI-related services cannot be provided in the EU market due to restrictions imposed by EU AI Act, in spite of restrictions being unspecified in its schedule. 

In such a situation, the EU should seek justification of market access restrictions under Article XIV GATS. 

The analysis of general exceptions under GATS Article XIV follows the same logic of two-tier test as under GATT Article XX, as WTO established in its case law [‎5, para. 291]. So, the steps of the test were already described with the occasion of arguing the eventual clash with GATT, and should be applied correspondingly.

To add, the EU could also justify the measure under Article XIV(c)(ii) if it is necessary to secure compliance with GATS-consistent laws or regulations relating to the prevention of deceptive and fraudulent practices; the protection of confidentiality and personal data of the individuals or safety. The use of AI is indissolubly connected with personal data and confidentiality of users, inevitably posing unacceptable and high risks enshrined in Articles 5 and 6 of the EU AI Act.  

Regarding security exceptions under GATS Article XIV bis, the reasoning is similar as under GATT Article XXI.

The TBT Agreement regulates the preparation, adoption and application of regulatory measures that include 1) technical regulation, 2) standards, and 3) conformity assessment procedures for assessing product compliance with such requirements [‎1, Annex I]. These measures can be used to achieve public policy goals, such as national security requirements; the prevention of deceptive practices; protection of human health or safety, animal or plant life or health, or the environment. At the same time, the TBT Agreement tends to ensure that these measures are not unjustifiably or arbitrarily discriminatory nor are they a disguised restriction on international trade [‎1, Article 13.1].

The EU demonstrates its openness to recognizing conformity-assessment results regardless of the jurisdiction they are established, aiming to respect in this way its obligations before WTO in light of avoiding unnecessary technical barriers in trade [‎17, Recital 127]. 

The provision indicates the EU AI Act, evidently, falls under the scope of the TBT Agreement. Scholars assume that the Act could be a technical regulation [‎13, p. 25]. This can be argued by the fact that the EU AI Act satisfies the three-tier test established in the Appellate Body’s interpretation of Annex 1.1 of the TBT in EC-Asbestos and EC-Sardines. Specifically, (i) the document must apply to an identifiable product or group of products, (ii) the document must lay down one or more characteristics of the product and (iii) compliance with the product characteristics must be mandatory [‎2, para. 66-70; ‎3, para. 176].

EU AI Act is a technical measure, since (i) it is a document that applies to products using AI technology, (ii) it lays down product characteristics based on their risk, and (iii) compliance is mandatory [‎13, p. 24].

Regarding the alignment with the TBT Agreement, it should be first noted that the EU AI Act reflects the legitimate objectives provided by Article 2.2 of the Agreement. This means the EU AI Act could initially be presumed as not creating an unnecessary obstacle to international trade. Nevertheless, the EU has discretion in the implementation of standards into technical regulations Therefore, a case-by-case analysis of the standard’s use would be necessary to determine compliance with Article 2.5 of the TBT Agreement [‎13, p. 36].

Another friction point is conformity assessment. Although Recital 127 of the EU AI Act does not apparently create unnecessary barriers to international trade, Article 39 seems to actually impose them. This provision authorizes the activity of conformity assessment bodies established under the law of a third country with which the EU has concluded a mutual recognition agreement (MRA). This clashes with TBT Agreement Article 6.1 on recognition of conformity assessment procedures’ results regardless of where and how they have been made, whenever possible. EU AI Act provisions let bypassing the TBT Agreement multilateral requirements.

This finding raises at the same time the discrimination issue under Article 2.1 of the TBT Agreement. The relevant EU AI Act provisions that might interfere with non-discrimination obligation is Article 43(1), along with Article 31(1). While Article 43(1) mandates third-party conformity assessment with the involvement of a notified body when harmonized standards are unavailable or not applied, Article 31(1) establishes, among others, that any such notified body must be established under the national law of an EU Member State and shall have legal personality.

To determine whether the EU AI Act conformity assessment regime is inconsistent with Article 2.1, the Panel should assess it through the following elements: (i) the measure at issue is a technical regulation withing the meaning of Annex 1.1 to the TBT Agreement; (ii) the relevant products are „like” products; (iii) the measure at issue accords less favourable treatment to imported products than to the relevant group of like products. 

The third element, in turn, includes two assessment steps: (i) determining whether the challenged measure modifies the conditions of competition to the detriment of the relevant imported products vis-à-vis like products of domestic origin and/or like products originating in any other country; and, if the panel makes such a finding, (ii) determining whether the detrimental impact on imports stems exclusively from a legitimate regulatory distinction (LDR) rather than reflecting discrimination against the group of imported products [‎6, para. 6.1-6.2].

Applying these norms to the EU AI Act mandatory third-party track for biometric systems, which are qualified as high-risk AI systems, reveals a potential discrimination for the following considerations. Assuming arguendo that the first two elements of Article 2.1 of the TBT Agreement are met, the assessment turns of the third element, less favorable treatment. 

Regarding the first step, because Article 31(1) of the EU AI Act requires the notified bodies to be established in the EU, and the EU lacks actual MRA on AI with third countries, foreign providers are barred from using domestically accredited conformity assessment bodies. Those providers will be forced to bear high compliance costs and efforts unlike the EU ones. This modifies the conditions of competition to the detriment of imported AI systems in the EU market. On the other hand, the EU might defend itself by invoking that he requirement to use EU-notified bodies reflects the legitimate objective of ensuring that assessment bodies are subject to EU supervisory oversight – auditable standards, harmonized procedures, and enforceable liability under EU law [‎9, p. 7]. 

As for the second step, the EU AI Act might say that notified bodies shall be situated on its territory, being a legitimate distinction necessary for close supervision and enforcement. Nevertheless, this argument does not automatically make the measure even-handed. Focus on geographic location of a notified body does not guarantee safety of AI systems per se. Also, not accepting foreign conformity assessment body findings does not bring benefits to the EU’s safety goals. Top AI-exporting countries, like the United States, India, South Korea, have their own accredited conformity assessment bodies that strive for high standards, not lower than EU ones.

From the practical perspective, the EU AI Act, prior to its adoption, became the very subject of specific trade concerns (STCs). On 11 November 2021, the EU notified the TBT Committee on the adoption of the Act [‎24].

In this context, between March 2022 and June 2023, the TBT Committee discussed trade concerns raised by China regarding the scope and meaning of the definition of AI systems; the fact that market surveillance authorities shall be granted access to the source code of the high-risk AI systems; ensuring proportionality of penalties and fines to the actual performance according to Article 71 of the EU AI Act; extending the transition period etc.

The EU addressed China’s concerns, arguing that the definition of AI should be technologically neutral, that the application of restrictions would be justified to protect the values enshrined in Article 1(1) of the EU AI Act, that the EU could grant access to source codes if two cumulative conditions are met, according to Article 74(13), that the penalties system is proportionate and so on.

To summarize, the EU AI Act is generally compatible with WTO disciplines, even though the treaties were drafted about 30 years ago without taking into account the rapid pace in technology development, including the emergence and exponential growth of AI. Although the EU AI Act norms might be excessively rigid in relation to WTO treaties, the EU does not automatically breach its obligations as WTO Member, the verdict on real interference shall be made on case-by-case basis.  

EU AI Act and its alignment with the SDGs

Beyond its impact on international trade, the EU AI Act serves as a pivotal regulatory instrument in aligning technological evolution with the UN 2030 Agenda on Sustainable Development [15]. The role of the EU AI Act for achieving relevant Sustainable Development Goals (SDGs) will be further argued.

The EU AI Act supports Goal 3 by classifying AI systems intended to be used for assessing the access and enjoyment by natural persons of essential private services, including healthcare ones, as high-risk (Article 6). By mandating strict conformity assessments and human oversight for AI in healthcare, the Act ensures that technological innovation does not compromise patient safety. This aligns with the WTO perspective that digital trade in healthcare must be supported by robust domestic regulation to ensure the delivery of high-quality, life-saving services. 

Obviously, the EU AI Act has the most convergences with Goal 8. The Act prohibits AI practices that use emotion recognition in the workplace (Article 5(1)(f)) and classifying AI used for recruitment or worker management as high-risk. These measures protect workers’ rights and dignity, preventing exploitative or biased automated management. From a trade perspective, the emphasis on ‘trustworthy AI’ creates a stable legal environment for digital services, which the WTO identifies as a prerequisite for sustainable economic growth and the expansion of digital trade.

Also, the EU AI Act contributes to growth of micro, small and medium-sized enterprises (MSMEs) and start-ups by reducing the compliance burden. The Act grants the SMEs and start-ups the right to provide the elements of technical documentation of high-risk AI systems in a simplified manner (Article 11(1)). Microenterprises, in turn, may comply with certain elements of the quality management system required by Article 17 of the EU AI Act in a simplified manner, provided that they do not have partner enterprises or linked enterprises within the meaning of the Commission Recommendation 2003/361/EC.

Regarding Goal 9, the EU AI Act encourages higher levels of economic productivity by providing a clear legal framework for fostering innovation. The inclusion of AI regulatory sandboxes specifically supports innovation by facilitating the development, training, testing and validation of innovative AI systems for a limited time before their being placed on the market or put into service pursuant to a specific sandbox plan agreed between the providers or prospective providers and the competent authority [‎17, Article 57(5)].

The EU AI Act is a critical tool for Goal 10 through its ban on AI-driven social scoring, in accordance with Article 5(1)(c), and the requirement for data governance to prevent algorithmic bias. By ensuring that high-risk systems are trained on representative datasets, the EU AI Act mitigates the risk of automating discrimination against marginalized groups and ensuring inclusive trade.

Finally, the achievement of the abovementioned SDGs would not be possible without supporting Goal 17. The EU AI Act aims to support promotion of a universal, rules-based, open, non-discriminatory and equitable multilateral trading system under the WTO, ensuring the implementation of Target 17.10. Additionally, the EU AI Act set up the AI Office, European Artificial Intelligence Board, and Advisory Forum for efficient governance at EU level. These bodies institutionalize the multi-stakeholder partnerships and international cooperation, facilitating the global exchange of technical expertise and policy coherence necessary to leverage AI for sustainable development. These bodies will integrate various stakeholders: industry, start-ups, SMEs, civil society and academia. This will ensure that the pursuit of technological progress remains inclusive, transparent, and aligned with the diverse interests of the global community.

Overall, the EU AI Act aligns with some of the SDGs, having the huge potential to contribute to the development of inclusive international trade, and preserving at the same time fundamental values of human rights and rule of law and democracy.

Conclusions

The EU AI Act represents the world’s comprehensive AI regulatory framework with substantial influence on international trade. The Act focuses not on the origin of AI systems or their deployers or providers, but persons located in the EU that can be affected by respective AI systems.

The issue of classification of AI as good or service was debated to determine the application of analyzed WTO disciplines: GATT, GATS, and the TBT Agreement. AI could be both good and service, even though the WTO Members have yet to reach an agreed solution.

If the AI system is placed on the market, made available on the market or put into service in the EU, and after the conformity assessment is categorized as a high-risk system or a prohibited one, it inevitably becomes subject to trade restrictions. Those should be justified either under general or security exceptions in light of GATT (if AI is incorporated in tangible goods) or GATS (if AI is an integral part of services).

The EU Act is convergent with the TBT Agreement, since it falls under the Agreement’s scope as a technical regulation. The Act explicitly provides the facilitation of mutual recognition of conformity assessment results regardless of the jurisdiction of the competent conformity assessment bodies. Also, AI Act’s purpose is in line with the legitimate objectives of the TBT Agreement, suggesting the presumption of avoidance of unnecessary barriers to trade. Nonetheless, the EU AI Act’s conformity assessment requirements may raise questions on non-discrimination and mutual recognition in practice. Concerns that might arise from these findings should be treated on case-by-case basis.

Lastly, the EU AI Act serves as the catalyst for the achievement of SDGs 3, 5, 9, 10, and 17, ensuring the effective and inclusive international trade for the benefit of all.

List of authorities:

1. Agreement on Technical Barriers to Trade, Apr. 15, 1994, Marrakesh Agreement Establishing the World Trade Organization, Annex 1A, 1868 U.N.T.S. 120, 33 I.L.M. 1173 (1994).
2. Appellate Body Report, European Communities – Measures Affecting Asbestos and Asbestos-Containing Products, WT/DS135/AB/R, adopted on 12 March 2001.
3. Appellate Body Report, European Communities – Trade Description of Sardines, WT/DS231/AB/R, adopted on 26 September 2002.
4. Appellate Body Report, United States – Import Prohibition of Certain Shrimp and Shrimp Products, WT/DS58/AB/R, adopted on 12 October 1998.
5. Appellate Body Report, United States – Measures Affecting the Cross-Border Supply of Gambling and Betting Services, WT/DS285/AB/R, adopted on 7 April 2005.
6. Appellate Body Report, United States – Measures Concerning the Importation, Marketing and Sale of Tuna and Tuna Products, WT/DS381/AB/RW/USA, WT/DS381/AB/RW2, adopted on 14 December 2018.
7. Commission, AI Act, https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai, accessed on 22 February 2026.
8. Commission, Proposal for a Regulation of the European Parliament and of the Council Laying Down Harmonised Rules on Artificial Intelligence (Artificial Intelligence Act) and Amending Certain Union Legislative Acts COM/2021/206 final.
9. Ehab Habila, Regulating the Algorithm as a Product: EU AI Act Conformity Assessment Requirements as Technical Regulations Under the WTO Agreement on Technical Barriers to Trade (2026). Available: http://dx.doi.org/10.2139/ssrn.6880583, accessed on 12 July 2026.
10. General Agreement on Tariffs and Trade 1994, Apr. 15, 1994, Marrakesh Agreement Establishing the World Trade Organization, Annex 1A, 1867 U.N.T.S. 187, 33 I.L.M. 1153 (1994).
11. General Agreement on Trade in Services, Apr. 15, 1994, Marrakesh Agreement Establishing the World Trade Organization, Annex 1B, 1869 U.N.T.S. 183, 33 I.L.M. 1167 (1994).
12. India AI Impact Summit Declaration, https://d19ob9sqegt2wc.cloudfront.net/stage/uploads/AI_Impact_Summit_Declaration_f208574dfc.pdf, accessed on 24 February 2026.
13. Milena da Fonseca Azevedo, From transparency to standards: the role of the TBT agreement in addressing AI regulatory challenges, Latin American Journal of European Studies 5, no. 2 (2025): 14-40. Available: https://eurolatinstudies.com/wp-content/uploads/2025/12/revistacompleta_v5n2_001.pdf, accessed on 22 February 2026.
14. Panel Report, Argentina – Measures Relating to Trade in Goods and Services, WT/DS453/R, adopted on 30 September 2015.
15. Panel Report, Brazil – Measures Affecting Imports of Retreaded Tyres, WT/DS332/R, adopted on 12 June 2007.
16. Panel Report, United States – Origin Marking Requirement, WT/DS597/R, adopted on 21 December 2022.
17. Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence and amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU) 2016/797 and (EU) 2020/1828 (Artificial Intelligence Act).
18. Takemasa Sekine, EUs AI Regulation and International Economic Law: The Complex Impact of the EU AI Act on Global Economic Governance, 2024, SSU Working Paper, University of Tokyo. Available: http://dx.doi.org/10.2139/ssrn.4997043, accessed on 22 February 2026.
19. UN General Assembly, Transforming our world: the 2030 Agenda for Sustainable Development (2015), UN Doc A/RES/70/1.
20. UN, CPC Version 3.0 Explanatory Notes (2025), https://unstats.un.org/unsd/classifications/Econ/Download/In%20Text/CPC_Ver_3.0_Exp_Notes_30Jun2025.pdf, accessed on 20 February 2026.
21. UN, Introduction to the Central Product Classification (CPC), Version 3 (2025), https://unstats.un.org/unsd/classifications/Econ/Download/In%20Text/CPC_Ver_3.0_Introduction.pdf, accessed on 20 February 2026.
22. UN, the Provisional Central Product Classification (1991).
23. WCO, Harmonized Commodity Description and Coding System, 7th edition, 2022.
24. WTO, Committee on Technical Barriers to Trade, Notification, EU, Artificial intelligence, G/TBT/N/EU/850.
25. WTO, Fourteenth Ministerial Conference – Declaration on Interim Arrangements for the Agreement on Electronic Commerce, WT/MIN(26)/W/26, adopted on 28 March 2026.
26. WTO, Fourteenth Ministerial Conference – MC14 Chairperson’s Summary, WT/MIN(26)/35, adopted on 31 March 2026.
27. WTO, Secretariat Services Sectoral Classification List (W/120), https://www.wto.org/english/tratop_e/serv_e/mtn_gns_w_120_e.doc.
28. WTO, Work Programme on Electronic Commerce, https://www.wto.org/english/tratop_e/ecom_e/ecom_work_programme_e.htm, accessed on 19 February 2026.
29. WTO, Work Programme on Electronic Commerce, Ministerial Decision, WT/MIN(24)/38, adopted on 2 March 2024.