[AI and New Civilization Standards Special Report] Economic Challenge ②: Challenges of AI and World Politics - Global South and Big Tech Issues

I. AI and the Global South

1. Problem Statement

Artificial Intelligence (AI) technology, spearheaded by Chat GPT, is bringing about changes in various social domains, including economics, military affairs, education, health, and culture, and its development is predicted to accelerate throughout the 21st century. Alongside the potential for AI to foster economic growth, various negative aspects are being discussed. This paper particularly focuses on the situation where the development, dissemination, and utilization of AI technology are led by developed countries, including the United States, and China, with the Global South being relatively marginalized.

The gap between developed and developing countries in the global political and economic order, the so-called North-South issue, has long been an important international political issue. Since the advent of the information age in the 21st century, numerous analyses have been made on the impact of IT technologies, such as the internet and smartphones, on the North-South gap (United Nations Trade and Development (UNCTAD) 2020). Research examining the medium- to long-term impacts of the rapid acceleration of automation, expansion of smart factories, and growth of online commerce markets, driven by COVID-19, reveals that digital transformation presents a complex picture with both positive and negative implications for the development of developing countries (Bae Young-ja 2022).

Developing countries generally have lower levels of digital infrastructure and technological innovation compared to developed countries, possess less capacity for digital transformation, and have fewer resources beyond labor. In this context, the digital transformation underway in traditional manufacturing sectors such as automobiles, particularly the spread of automation and smart factories, is likely to replace the relatively simple and repetitive labor that developing countries have performed within global value chains with machines. This scenario suggests a higher probability of a reduction, rather than an enhancement, of developing countries' standing, and a potential widening of the gap between developing and developed countries. On the other hand, there are aspects where digital transformation in developed countries can also present opportunities for growth in developing countries. Indeed, some regions in developing countries, such as Mexico, have benefited from increased trade with developed countries by participating more actively in global value chains. The spread of smart factories utilizing robots and 3D technology in developed countries has led to increased productivity and imports, which in turn has boosted demand for final products, thereby stimulating trade with developing countries that supply parts and intermediate goods (UNCTAD 2020).

Regarding online commerce, it is true that the spread of digital devices and the internet has increased opportunities for developing country businesses to participate in global value chains as suppliers and producers through online marketplaces. It has become more common for small and medium-sized producers in developing countries to utilize digital technologies to trade with consumers in developed countries through major e-commerce sites such as Alibaba, Amazon, eBay, Taobao, and Mercado Libre. However, the transaction methods, delivery systems, payment systems, and refund/exchange policies required by these platforms pose entry barriers for businesses in developing countries. Furthermore, the data monopolies of online commerce platform companies and their anti-competitive practices, such as predatory pricing and demands for high commissions, are exacerbating the difficulties faced by small-scale producers in developing countries. In other words, while digital transformation can serve as an opportunity for the development of developing countries, there are significant aspects that negatively impact their development, including the platformization of e-commerce companies, data monopolies, changes in value chains, and a reduction in employment effects.

Amidst the ongoing trend of digital transformation, how is the development and utilization of AI technology, which has been accelerating recently, impacting the gap between developed and developing countries? Will AI further amplify existing disparities? Or will AI serve as an opportunity for developing countries, contributing to reducing the gap? What crises may arise in the long term if developing countries are excluded from AI development and utilization? This paper seeks to explore ways to enhance the global political and economic standing of the Global South by examining the differences in AI infrastructure investment and utilization between developed and developing countries, and by considering the discourse surrounding AI and developing countries, while minimizing the negative impacts and maximizing the positive impacts of AI on the development of developing countries.

2. Status of AI Infrastructure, Technology Investment, and Utilization: A Perspective on the Gap Between Developed and Developing Countries

We will gauge the extent of the gap between developed and developing countries in the development, dissemination, and utilization of AI through various indicators. First, let us examine internet penetration rates, a fundamental infrastructure indicator for running AI. As shown in Figure 1 below, out of the current world population of approximately 8 billion, about 2.7 billion people do not have internet access. Notably, 680 million people in India and 340 million in China are reported to be without internet access. Furthermore, internet penetration rates in Sub-Saharan Africa are below 40 percent, and in Southeast Asia, they are as low as 51 percent, indicating a significantly low level of access.

<Figure 1> Internet Penetration Rate

Source: Datareportal. 2024

The investment in artificial intelligence, which has been accelerating since 2020, is also led by a few developed countries. According to Figure 2, from 2013 to 2023, the United States led AI-related investments with $330 billion, followed by China with approximately $100 billion. Other countries such as the United Kingdom, Israel, Canada, and Germany have invested tens of billions of dollars. The investment by the United States is so dominant that the combined investment from China (ranked second) to Sweden (ranked fifteenth) barely exceeds half of the US investment.

<Figure 2> AI Investment by Country (2013-23)

Source: HAI, Stanford. 2024.

The lack of AI-related infrastructure and investment in developing countries naturally leads to a low output of AI research and patents. In AI research, the United States, China, and EU countries are leading. According to Figure 3, from 2014 to 2023, US research institutions produced over 772,000 papers in AI research (peer-reviewed papers in core AI fields), accounting for a 30% share. China produced approximately 465,000 papers, with an 18% share. Other AI research hubs include the UK, Germany, Japan, India, Brazil, and Iran, each producing between 10,000 and 140,000 papers. In contrast, AI research in Africa, South America, and most Asian countries accounts for less than 5%. In terms of AI patents from 2010 to 2022, South Asian, Latin American, and Sub-Saharan African countries accounted for less than 1%, with most patents granted to China (61.1%) and the United States (20.9%).

<Figure 3> Share of AI Research and Patents

Source: Digital Science. 2024.

Foundation models, such as GPT-4, Claude 3, and Llama 2, which are machine learning or deep learning models trained on vast amounts of data to be applicable to a wide range of cases, are driving the spread of AI. Approximately 150 such models have been released, representing AI research and utilization. By nationality, the United States has 109, China has 20, and the United Kingdom has 8. The UAE, Canada, Singapore, Israel, Germany, and Finland each have two.

Interestingly, despite their significant weaknesses in AI infrastructure, investment, research, and patents compared to developed countries, developing countries show relatively high awareness and utilization of ChatGPT. According to data, the proportion of respondents using ChatGPT daily or weekly in India, Kenya, and Pakistan is 75%, 69%, and 62%, respectively, significantly exceeding the global average. This is higher than in Germany (41%), Japan (38%), and the UK (38%). This can be interpreted in various ways, but key factors include the aggressive market entry of Big Tech companies and the expansion of AI services to less developed countries at a relatively low cost in the health and education sectors. Big Tech companies like Google, Microsoft, and Meta have expanded by establishing R&D centers and engineering offices in developing countries (Okolo 2023). IBM established IBM Research India in 1998 and subsequently set up research labs in São Paulo and Rio de Janeiro (2010), Nairobi (2013), and Johannesburg (2016). In 2005, Microsoft opened Microsoft Research India in Bangalore and established two African development centers in Nairobi and Lagos. Google established AI research labs in Accra in 2018 and Bangalore in 2019.

Crucially, despite their vulnerability in terms of infrastructure investment performance, the spread and utilization of AI are progressing more actively than expected in developing countries. Furthermore, while the adoption rate of generative AI by companies and institutions is 40% in the United States, it is 31% in China and 33% in developing countries, including India and Latin America. Considering AI utilization, there remains potential and room for AI to become an opportunity for developing countries.

<Figure 4> Global ChatGPT Usage Frequency (2023)

Source: Stanford University AI Index Report. 2024.

According to UNCTAD reports, the gap between developed and developing countries has widened since the start of the information age in the 1980s. As observed above, developed countries lead in all aspects, including AI infrastructure investment, research investment, and foundation models, with a very low presence of developing countries. Therefore, it is difficult to be optimistic that the currently accelerating AI technology development and utilization will lead to a trend of narrowing the existing gap. PwC estimates that AI could contribute up to $15.7 trillion to global economic growth by 2030, with the economic impact on the Global South, excluding China, estimated at only $1.7 trillion. However, given the active adoption and utilization of AI in less developed countries, we must actively seek ways to reduce the gap while observing the overall impact of AI on the development of these nations.

<Figure 5> Historical Trend of the Developed-Developing Country Gap

Source: UNCTAD, Technology and Innovation Report. 2023.

3. Discourse on AI and the Global South

1) Algorithmic Imperialism

The discourse of so-called algorithmic imperialism has emerged, suggesting that the dominance of AI by the West not only leads to the marginalization of less developed countries in technological advancement but also extends to the dominance and control of social, political, and cultural discourse (Birhane 2023). According to this logic, while traditional colonialism was state-led, algorithmic colonialism is corporate-led. The former took the form of indiscriminate violent domination, whereas colonialism in the AI era differs in that it takes the form of 'cutting-edge algorithms' and 'AI-driven solutions' to social problems. Much of Africa's digital infrastructure and ecosystem is predominantly operated by Western companies such as Google, Meta, Netflix, and Uber. For instance, in the process of creating population maps of Africa based on data obtained through Meta's computer vision technology, demographic data, and high-resolution satellite imagery, Meta naturally gains the authority to generate and control knowledge about the continent's population. For data extraction and profit maximization, Meta not only freely utilizes information about Africans and their movements but also authoritatively determines what constitutes legitimate knowledge about the population. This is very similar to how Westerners in the past claimed to know better what the colonized needed, that they would save them, and that the colonized should be grateful to them, in order to justify colonial rule, with only the technology used having changed.

Nigeria is relatively technologically advanced in Africa, yet approximately 90% of the software used there is imported, creating a barrier to indigenous technological development. AI developed in the West is not only unsuitable for African problems but also, the intrusion of Western algorithms hinders the development of locally relevant products while making the continent dependent on Western software and infrastructure. For example, AI diagnosis using mammotomes, which is effective for breast cancer diagnosis in the West, has not yielded expected results in Africa, with self-diagnosis reportedly being more useful. This suggests that health-related AI systems developed based on Western lifestyles and environments may have limited utility in less developed countries and should be used with caution.

Due to biases inherent in AI models and data, problems are often framed and solutions derived in ways that align with Western perspectives. For instance, it is reported that when the same question is posed to AI models, the resulting answers are biased towards the West, unlike the results obtained offline from sources like the World Values Survey (WVS) and the Pew Research Center. In a similar vein, the argument is made that AI systems monopolized by developed country corporations will ultimately monopolize and control knowledge, leading to 'knowledge slavery' for developing countries (UNCTAD 2023).

<Figure 6> Comparison of Survey Results (2023, HAI)

Source: HAI. 2023.

The dominance of Western AI infrastructure and utilization extends beyond mere technological and data monopolies to encompass the spread of Western perspectives and lifestyles, leading to the uncritical acceptance of the Western model's legitimacy. The adoption of AI models that disregard the political, economic, and cultural contexts of less developed countries often proves unsuitable for solving their pressing issues and is more likely to deepen their dependence on infrastructure and data.

2) Discourse of Opportunity

On the other hand, there is also the argument that AI serves as a useful tool for solving various problems faced by developing countries and can enhance democracy by encouraging citizen participation (Okolo 2023). These proponents view AI as an opportunity for the development of developing countries and actively recommend its utilization. Over the past decade, less developed countries have begun using AI tools to address traditional development challenges, with increasing applications in agriculture, healthcare, and education. For example, deep learning models for field diagnosis of banana diseases and crop infections have been developed to support farmers in developing countries, along with image observation systems to aid in agricultural and forestry monitoring. In healthcare and education, services have been developed and are being piloted to compensate for infrastructure shortages and drastically reduce costs using AI. In India, predictive models have been built to ensure that pregnant women in remote areas can continuously participate in remote healthcare support programs, and a clinical decision support system has been created in Ghana to combat antibiotic resistance. In education, Colombia uses AI to identify students at various risks, Thailand enhances English language learning for its students with AI, and AI assistants are utilized to support science education in West Africa.

They argue that current AI practices can be democratized, more inclusive AI systems can be encouraged, and the participation of underrepresented communities in AI development can be increased. There is a growing demand for regional specialization and development of AI systems, with active efforts underway in this area. Grassroots civic groups such as Masakhane and Ghana NLP are attempting to develop datasets and machine translation tools that help expand accessibility to African languages. Additionally, organizations like Deep Learning Indaba, Khipu, AI Saturdays Lagos, and Data Science Africa are contributing to building local expertise in AI and fostering the growth of AI researcher and developer communities in Africa and Latin America.

4. Crises Caused by the AI Gap and Response Measures

What situation will the world face if AI development and utilization are concentrated among a few developed countries, excluding developing nations? Specifically, what crises can be predicted? In Table 1, we have categorized and summarized the various crises that can be considered, along with their content and response measures, into economic disparities between developed and developing countries, data and technological dependence of developing countries, expansion of socio-cultural inequality and bias/discrimination, and military and political instability.

<Table 1> Crises and Response Measures Resulting from the Widening AI Gap Between Developed and Developing Countries

To address the crisis stemming from the widening AI gap between developed and developing countries, efforts at both the individual country level (developed and developing) are important, but international efforts to recognize and respond to this crisis are essential. An AI global governance framework that mitigates the AI gap issue must be established.

5. AI Global Governance and the Global South

While AI technology has the potential to reinforce algorithmic colonialism, it also offers significant opportunities for active utilization in sectors such as agriculture, health, and education, particularly in developing countries. However, challenges remain for these nations to effectively leverage the benefits offered by AI. Despite steady increases in electricity and internet penetration in Africa, a significant portion of the population in sub-Saharan Africa still lacks internet access. The World Bank estimates that at least $100 billion in investment is needed to connect 100 million unconnected Africans in remote areas. Infrastructure expansion in Africa is crucial. The 2Africa cable, the longest subsea internet cable ever deployed, is currently being rolled out. Most of the estimated 485 subsea internet cables in use worldwide are owned by major telecommunications companies. Big tech companies, including Amazon, Google, Meta, and Microsoft, are also increasing their investment in subsea infrastructure, jointly owning approximately 30 cables. Discussions are needed on how this infrastructure expansion can consider the interests of underdeveloped regions.

Developing countries are responsible for labor-intensive data labeling, and the exploitation of data workers and content moderators in East Africa and South Asia, employed by companies like Sama and Scale AI, has been a point of concern. Furthermore, the lack of data protection and AI policies in most developing countries potentially leads to more frequent data misuse. Currently, concerns and discussions about the negative impacts of AI are primarily focused on developed countries, necessitating active attention and discussion regarding the situation in developing countries.

More proactive efforts from governments in developing countries are also important. To foster the expansion of their AI ecosystems, these governments should build capacity for training local researchers and developers through partnerships with external organizations, and support the creation and maintenance of AI ecosystems that encourage entrepreneurship and foster local innovation. Nigeria has partnered with Microsoft to provide digital skills to its citizens, and Google has been educating approximately 8 million people in Latin America in digital skills since 2017. Plans to integrate topics such as computer literacy, ICT skills, coding, digital citizenship, and online safety into K-12 curricula are mentioned in digital innovation initiatives in Brazil, Costa Rica, India, Jamaica, Malaysia, Panama, Rwanda, and South Africa.

As the influence of AI technology grows, it is crucial to create an environment where developing countries have fair access to AI technology, enabling equitable competition and utilization. Various international organizations, including the UN, World Bank, IMF, and G20, advocate for the promotion of AI to ensure the stability, security, and sovereignty of developing countries while advancing human interests. Ideas such as a digital public infrastructure system framework and the Global Digital Public Infrastructure Repository (GDPIR) have also been proposed. The G20's Data Governance Initiative 3 (DGI-3) supports developing countries in operating AI models using public datasets in the climate change domain.

While various international organizations and conferences related to AI technology and its diffusion are continuously discussing the North-South AI gap, mitigation measures, and support for developing countries, it is requested that developing countries be included and invited to participate in AI global governance. Through this, awareness of the crises that the AI gap may bring should be raised, and measures to respond to them should be sought.

II. AI and Big Tech Companies

1. AI Innovation and Regulation

Big tech companies, such as Google, Amazon, Apple, Meta, and Microsoft, which own or control large digital platforms, provide monopolistic services based on network effects. They also collect and accumulate real-time data crucial for understanding consumer preferences and predicting behavior, forming the core of the digital ecosystem. Google operates in over 200 countries, Meta has 2.3 billion monthly active users worldwide, and MS's Office and Windows have over 1.2 billion and 1.4 billion users, respectively. Based on their monopolistic positions, they have secured immense financial resources, accounting for 22% of the S&P 500 market capitalization, with individual companies exceeding the GDP of Canada and Italy.

The emergence of generative AI further strengthens the dominance of big tech. Developing generative AI requires significant human resources, data, and computing power. Few companies beyond big tech possess the tens of thousands of top-tier professionals, vast amounts of data, and the computing capacity to process data rapidly, enabling the purchase of over ten thousand GPUs. As shown in <Figure 7>, current generative AI operations are dominated by a small number of big tech companies.

<Figure 7> Global Generative AI Market Share

Source: IoT Analytics Research 2023-Generative AI Market Report. 2023-2030.

A critical issue related to AI concerns how to balance AI technological innovation with regulation. While competition among big tech companies accelerates AI innovation and diffusion, it also raises concerns about associated risks and negative aspects. As AI technology can bring benefits to various societal domains such as transportation, health, education, and production, countries are actively developing AI strategies and supporting corporate innovation. On the other hand, discussions are actively underway regarding how to minimize and regulate the risks and problems caused by AI. The most urgent issue is how to prevent and regulate the development of harmful or unethical AI systems, systems with inherent bias or discrimination, or systems that can be used for malicious purposes, such as cyber warfare. Furthermore, finding solutions is important because intense competition among big tech companies and monopolies by a few companies can ultimately stifle innovation and limit the potential benefits of AI. AI global governance is needed to ensure that AI is developed and used responsibly and ethically, and that the development, utilization, and effects of AI technology are shared by all.

2. Expanding Role of Big Tech Companies and Their Relationship with Governments

In the AI era, big tech companies are deeply involved not only in economic influence but also in agenda setting, providing discourse for problem-solving, and the policy implementation process based on that influence. Big tech plays a pivotal role in identifying problems and setting priorities. For example, they can effectively set agendas by utilizing online content for issues under discussion, and they participate in academic and policy research to explore and seek solutions for these problems. Big tech influences research by conducting it directly, funding academic and research institutions, or through researcher rotations. The frequency and proportion of big tech conducting their own research is increasing, and they raise new policy issues by analyzing vast amounts of data. Similar to how Amazon owns The Washington Post or Apple launched Apple News, big tech is attempting to transform from content moderators to content providers by directly owning media outlets.

In the past, essential infrastructure such as railways and roads were owned or operated by the government. However, new information infrastructure, including search engines and browsers, data servers and cloud computing, email and instant messaging, social networking, app stores, payment systems, video hosting, geographic information, and navigation services, are entirely owned by big tech companies. Governments are compelled to seek the assistance of big tech when facing specific challenges, such as during the COVID-19 pandemic. In fact, many governments invited representatives from big tech to design policies for managing COVID-19. With their cooperation, they were able to create epidemic spread maps, design tracking tools, formulate disease prevention policies, and distribute medicines and vaccines. Collaboration and support from big tech are becoming essential elements in increasingly diverse policy areas. As the economic and social influence of big tech companies grows, so does the demand for their social responsibility, and big tech companies are participating more directly as providers of policy services. Support activities from big tech are actively underway in the health, education, and finance sectors of developing countries.

With the advent of generative AI, big tech companies are moving beyond agenda setting and redefining societal problems to proposing solutions in the policy-making stage and even taking on the role of decision-makers. As big tech has historically accumulated cross-border profits on an unprecedented scale and gained immense political influence, they are sometimes treated as actors with a status akin to sovereignty. For instance, the Danish government has appointed a Tech Ambassador based in Silicon Valley to engage with big tech and represent Denmark's interests, indicating a perception of big tech companies holding a status similar to that of a nation.

As big tech companies are increasingly recognized as quasi-governmental or near-state entities, attention is focused on how the relationship between states and big tech will be configured. As shown in <Figure 8>, the relationship between the two is expected to evolve through various pathways. Firstly, competition and direct conflict between sovereign states and big tech are inevitable in certain areas. Conflicts may arise, particularly when governments strengthen regulations on big tech and big tech pushes back (competition). On the other hand, when governments seek to strongly regulate big tech, big tech may adopt a strategy of avoiding direct conflict with the state and aligning with government policies as much as possible (cooption). Conversely, if governments do not strengthen regulations on big tech and adopt a laissez-faire approach based on free market principles, the relationship between big tech and the government can become highly cooperative (collaboration). As the influence of big tech grows, there are instances where governments may become captured by big tech (capture). The four directions are not mutually exclusive and may overlap.

<Figure 8> Relationship Matrix between Governments and Big Tech Companies

Source: Khanal et al. 2024.

3. Crises and Response Measures Arising from Big Tech-Led AI Development and Utilization

We have organized the anticipated crises and response measures when big tech leads AI development into categories such as market monopoly and competition restriction, labor market changes and unemployment and economic inequality, data bias and privacy infringement, technological control and threats to democracy, challenges to state power by big tech, and algorithmic imperialism by big tech, as presented in <Table 2>.

<Table 2> Crises and Response Measures Caused by Big Tech-Led AI Technology Development

4. Current Status of AI Norms

As is well known, generative AI is expected to contribute to increased corporate productivity by enabling cost reductions and data-driven decision-making. Generative AI is also fostering the emergence of new services in areas such as text, image, and music creation. On the other hand, generative AI learns from vast amounts of data but can sometimes produce inaccurate or biased responses due to data limitations. The data used by generative AI may include sensitive personal or private information, posing a risk of privacy infringement. Generative AI acquires data from existing copyrighted works, thus carrying a high possibility of intellectual property infringement. The use of generative AI can facilitate the spread of misinformation and manipulation of public opinion.

The various inherent risks of AI technology can be amplified when big tech companies lead its development and utilization. This is because, amidst fierce competition for market share and accelerated investment among big tech firms, the norm of developing trustworthy, responsible, and ethical AI technology is difficult to adhere to as the pursuit of dominance takes precedence. Currently, discussions regarding big tech primarily occur within the framework of antitrust regulation at the domestic level. At the international level, there is a broad consensus that AI global governance is needed to establish ethical norms that can address the negative aspects of AI technology and implement them. However, given the slight differences in AI regulation stances among the US, Europe, and China, it is difficult to be optimistic about the emergence of agreed-upon international norms and effective AI global governance.

The EU has established the Artificial Intelligence Act (AIA) in 2024, becoming the first to lay the legal groundwork for comprehensive AI regulation. The AIA aims to promote the use of human-centric and trustworthy AI while minimizing the harmful effects of AI to protect health, safety, fundamental rights, and democracy. The EU AIA categorizes AI systems based on their risk level to users into unacceptable risk, high risk, limited risk, and minimal risk, applying different levels of regulation accordingly. Specifically, systems that ① manipulate human decision-making using subconscious or deceptive techniques, ② exploit the vulnerability of individuals or specific groups, ③ treat individuals or specific groups unfairly, or evaluate or classify social scores, or ④ utilize real-time remote biometric identification technology in publicly accessible spaces are designated as unacceptable and their use is fundamentally prohibited. AI technologies used in education, healthcare, and military sectors that could pose risks to fundamental rights must meet the new standards set by the EU, and companies must transparently disclose their model development methods and be held accountable for damages. The EU is pressuring big tech companies like OpenAI (GPT-4) and Google (Gemini) by stating that their AI models pose risks and require additional work to meet EU standards.

The background for the EU's pioneering legislation on AI regulation and its leadership in shaping international norms lies in its intention to check the activities of US big tech companies within Europe and lead safety-related standards, given that Europe itself lacks large-scale digital platform companies. While not leading in AI technology, the EU is actively responding to AI safety regulations and evaluations. Furthermore, there is an expectation of establishing safety evaluation services and institutions within Europe by leading the formation of international standards for AI safety.

In the United States, the general stance is that corporate self-regulation is appropriate for AI safety, with the government announcing policies to complement this self-regulation. Most big tech companies have established and published their own ethical guidelines, but no one believes these are sufficient on their own. Following the release of the Federal AI Bill of Rights in 2022, the Biden administration signed the Executive Order on Safe, Secure, and Trustworthy Artificial Intelligence in 2023. This executive order mandates AI safety assessments, establishes safety standards for AI tools, creates content authentication standards, and enhances privacy protection. The core objective is to promote the safe and responsible development and use of AI at the federal level while regulating AI technologies that threaten national security, health, and safety. The National Institute of Standards and Technology (NIST) will establish the US AI Safety Institute (US AISI) to operate an AI risk management framework, including guidelines, tools, and benchmarking, for evaluating and mitigating AI risks.

The U.S. Department of Homeland Security established the AI Safety and Security Board in 2024, notably including CEOs from 22 major domestic AI companies such as OpenAI, Anthropic, Microsoft, Google, Nvidia, IBM, Adobe, and AWS as advisory members. This demonstrates the U.S. government's intention to seek ways to enhance AI technology's safety and security by collaborating with big tech and leveraging their expertise, rather than unilaterally regulating them.

Recently, the EU's enactment of the AI Act has imposed obligations on many US companies to disclose content used in AI training, comply with mandatory risk prevention measures, and has restricted the use of real-time remote biometric identification systems. In response, big tech companies are exploring countermeasures, and the US government is also making efforts not to lose its leadership in AI regulation to Europe. The US submitted a resolution to the UN in May 2024, "Seizing the opportunities of safe, secure and trustworthy artificial intelligence systems for sustainable development," which was approved with 143 votes in favor, 9 against, and 25 abstentions. The resolution emphasizes the harmful impact of inappropriate or malicious AI use on human rights and the risks of increased inequality and discrimination due to biased data, calling for the coordination of global norms, but it is non-binding. The resolution specifically calls for international cooperation and global consensus for the safe development and implementation of AI, cooperation and sharing of research and technology among stakeholders, the development of regulations at national and international levels, the prohibition of misuse in AI design, development, deployment, and use, and advocates for technology transfer and financial support to bridge the digital divide between developed and developing countries. The US fundamentally acknowledges the self-regulation of its domestic big tech companies and implements complementary policies, while simultaneously raising the need for international norms through the UN to counter the EU's unilateral leadership in AI international standards.

China is also pursuing AI trustworthiness and establishing certification systems through a series of principles and guidelines. In May 2019, it released the 'Beijing AI Principles.' These principles outline norms for the development, use, and governance of AI, with fundamental values of pursuing human well-being and diversity and inclusion. For AI development, seven principles were presented: utilization for humanity, pursuit of the right goals, accountability, risk controllability, ethics and transparency, diversity and inclusion, and openness. For AI use, three principles were presented: responsibility for use, human rights and data management, and education and training. For AI governance, five principles were presented: employment optimization, harmony and cooperation in the AI ecosystem, appropriate regulation, segmented guidelines, and long-term strategy development. Internationally, at the 2023 Belt and Road Forum, it proposed a Global AI Governance Initiative, urging equal rights and opportunities for all countries in AI development and opposing technological monopolies and unilateral coercive measures.

China emphasizes equal rights and opportunities for all countries while countering the technological monopoly of US big tech, presenting guidelines for its domestic AI companies to establish norms and further emphasizing support for its own AI companies. Following criticism for the indiscriminate use of facial recognition technology by Chinese companies, China released draft regulations for the safe management of facial recognition technology applications. Recognizing the urgent need to block the spread of false information, bias, and the potential for public opinion manipulation by generative AI, China has implemented interim measures for the management of generative artificial intelligence services. This indicates that AI norms or regulations in China tend to be somewhat defensive.

In the UK, the first AI Safety Summit was held at Bletchley Park in 2023, where the Bletchley Declaration was adopted, calling for a joint response to the potential risks of artificial intelligence. The core agenda of the summit was to seek and establish consensus on swift and appropriate regulatory measures to ensure human-centric, trustworthy, and responsible AI. The summit was attended by 28 countries, including South Korea (Australia, Brazil, Canada, Chile, China, European Union, France, Germany, India, Indonesia, Ireland, Israel, Italy, Japan, Kenya, Saudi Arabia, Netherlands, Nigeria, Philippines, Rwanda, Singapore, Switzerland, Türkiye, Ukraine, UAE, United Kingdom, United States), which pledged to cooperate on AI safety. By leading international conferences on AI safety involving not only the US and China but also developing countries, the UK is striving to counter the AI technology development and diffusion led by the US and China and to form a more representative and robust international solidarity.

Due to the intense competition among nations surrounding AI technology and their somewhat differing stances on norms, difficulties are anticipated in the emergence of internationally agreed-upon norms or governance. However, there is a growing consensus on the need for global AI governance to ensure AI safety and trustworthiness, manage them, and regulate big tech. Discussions are underway regarding which existing international organizations would be most suitable for the AI domain. A recent report compares examples such as the International Civil Aviation Organization (ICAO) for civil aviation, CERN and the International Atomic Energy Agency (IAEA) for nuclear and atomic energy, the Intergovernmental Panel on Climate Change (IPCC) for climate change, and existing international organizations for managing international payment systems (The Bank for International Settlements: BIS; Financial Stability Board: FSB; Financial Action Task Force: FATF). That is, international organizations capable of exchanging information, promoting cooperation, monitoring, and enforcing, such as those managing nuclear energy, civil aviation, and international payment systems, are proposed as suitable candidates for global AI governance.

<Figure 9> Comparison of Global AI Governance Types

Source: Microsoft. 2024.

To date, a small number of advanced countries' big tech companies have led the development and utilization of AI technology, raising concerns domestically about the monopolization by big tech and internationally about the widening gap between developed and developing countries. These circumstances, combined with the growing awareness of the safety and reliability risks posed by AI technology, are leading countries to establish regulations on AI technology, and global discussions on AI norms and governance are also underway. The pace of AI technology development and utilization far outstrips these discussions, failing to reduce the uncertainty surrounding AI technology. As competition and differences in positions among major countries regarding AI norms become apparent, the future direction of AI norms and governance is difficult to be optimistic about. However, as the continuous acceleration of AI technology development is predicted, issues of ethics, reliability, and safety will become even more acute, and the gap created by AI technology will also emerge as a significant challenge that the international community must address together. The direction of AI technology development and governance will be explored through the dynamic interactions of fierce competition among big tech companies vying for leadership in AI technology development, the US government's efforts to establish AI safety and accountability norms within the scope of supporting its domestic big tech companies' advantages, the EU and UK governments' moves to lead international efforts in establishing AI norms while checking US big tech companies, China's response by actively supporting its domestic AI companies and imposing regulations such as facial recognition and generative AI management guidelines, and the Global South's position, which criticizes the AI technology and norm-setting efforts led by developed countries and requests attention to the AI gap. ■

References

Bae, Young-ja. 2022. "COVID-19 and the Digital Economy: Perspectives on Global Value Chains and Developing Country Development." EAI Working Paper. February 9.https://www.eai.or.kr/...kor_workingpaper.

Birhane, Abeba. 2020. "Algorithmic Colonization of Africa." SCRIPTed 17, 2: 389.https://script-ed.org/?p=3888.

Draux, Hélène. 2024. "Research on Artificial Intelligence – the Global Divides." Digital Science, January 4.https://www.digital-science.com/...divides/.

Huw, Roberts, Emmie Hine, Mariarosaria Taddeo, and Luciano Floridi. 2024. "Global AI Governance: Barriers and Pathways Forward."International Affairs 100, 3: 1275-1286.https://doi.org/10.1093/ia/iiae073.

Kemp, Simon. 2024. "Internet Use in 2024." Datareportal, January 31.https://datareportal.com/...adoption

Khanal, Shaleen, Hongzhou Zhang, and Araz Taeihagh. 2024. "Why and How Is the Power of Big Tech Increasing in the Policy Process? The Case of Generative AI." Policy and Society, puae012.https://doi.org/10.1093/polsoc/puae012.

Smith, Brad, and Natasha Crampton. 2024. "Global Governance: Goals and Lessons for AI." Microsoft, May 17.https://blogs.microsoft.com/...ai/.

Stanford Human-Centered Artificial Intelligence (HAI). 2024. “Artificial Intelligence Index Report 2024.”https://aiindex.stanford.edu/...Report.pdf.

Maslej, Nestor, Loredana Fattorini, Raymond Perrault, Vanessa Parli, Anka Reuel, Erik Brynjolfsson, John Etchemendy, Katrina Ligett, Terah Lyons, James Manyika, Juan Carlos Niebles, Yoav Shoham, Russell Wald, and Jack Clark. 2024. "The AI Index 2024 Annual Report." AI Index Steering Committee, Institute for Human-Centered AI, Stanford University, April.https://aiindex.stanford.edu/report/.

Okolo, Chinasa. 2023. "AI in the Global South: Opportunities and Challenges Towards More Inclusive Governance." Brookings, November 1.https://www.brookings.edu/...governance/.

UN Trade and Development (UNCTAD). 2023. “Technology and Innovation Report 2023.”https://unctad.org/tir2023.

_________. 2020. “World Investment Report 2020.”https://unctad.org/...2020.

World Bank. 2020. “World Development Report 2020: Trading for Development in the Age of Global Value Chains.”https://www.worldbank.org/...wdr2020.

■Bae Young-ja, Professor of Political Science and International Relations, Konkuk University.

■ Responsible Editor and Publisher:Park Ji-soo, EAI Researcher

    Inquiries and Editorial: 02 2277 1683 (ext. 208) | jspark@eai.or.kr