An Open Door

Artificial intelligence (AI) is progressing at remarkable speed, with some experts estimating that the technology will be performing many thousands of times better in 2028 than it was at the start of 2025. Thought of as a machine system that can make “predictions, recommendations, or decisions” based on a certain set of parameters with human-like reasoning, AI is not a new technology, but recent developments in generative AI, which can create new content by mimicking human speech and interaction, has captured widespread attention. Recognizing the very real impact that this technology has—and will have—on every facet of society, governments and the private sector have increasingly invested billions of dollars in conducting research and development (R&D), building computing capacity, and attracting the best talent.

The economic impact of AI reaches into the trillions: According to the International Data Corporation, AI will add $19.9 trillion to the global economy by 2030. However, these impacts are unlikely to benefit the world equally, and already there is a growing digital divide between the United States, China, and Europe on the one hand, and the rest of the world on the other. If current development trends continue, it is expected that only 3 percent of the projected AI economic benefits will go to Latin America, 6 percent to developed countries in Asia (excluding China), and a mere 8 percent to the combined populations of “Africa, Oceania, and other Asian markets” (again, excluding China).

Given the economic benefits that AI will bring to those countries that have the ability to harness it, recent progress has spurred heated debate on several related fronts, including

• the ramifications of only a few developed countries dictating AI development;

• the opportunities for open-source AI technologies to help close the digital divide;

• the associated risks that open-source AI poses for national security; and

• the impact that great power competition between the United States and China could have on future innovation.

AI principles commonly promoted by G7 countries, such as transparency, accountability, fairness, and explainability, are currently based on assumptions around levels of agency unique to the Global North and neglect to view the Global South as a partner in AI’s development. There are additional concerns over AI bias, since, to date, the technology has predominately reflected the views, languages, and ideologies of select advanced economies.

Developing countries recognize that without the ability to adapt current AI models or build new models to best fit their needs, they risk being trapped in a new cycle of digital colonialism. As they look for partnerships to build their own AI ecosystems, they are increasingly faced with a choice between the United States and China. Many AI breakthroughs have come from U.S.-based innovation, putting the United States in a position to potentially forge rewarding partnerships with countries in the Global South looking to leverage AI for sustainable development. However, the spillover effect of U.S. export controls aimed at stemming Chinese AI progress could also threaten the abilities of developing country partners to build their own systems and could risk turning them toward China in a bid to access these technologies. This paper seeks to unpack these options within the context of great power competition, and their implications for developing countries’ ability to harness the power of AI for their own development. The paper also examines the role that open AI models could play in lowering the barrier to entry for developing countries; current AI development trends across low- and middle-income countries (LMICs); and ways in which the United States can strengthen its tech partnerships to ensure that the next era of AI remains oriented toward principles of transparency, democracy, and safety.

AI Development and Governance on the Global Stage

A handful of countries have primarily led the development of AI, and among them, the United States and China are at the forefront. In 2024, U.S. private sector investment in AI reached $109.1 billion, and at time of writing, 100 firms, mainly located in the United States and China, account for roughly 40 percent of corporate R&D spending on AI. The market values of each of the largest tech companies—Apple, Nvidia, and Microsoft—are roughly equal to Africa’s GDP. As the money suggests, the countries with the ability to invest in AI are doing so—at rates developing economies cannot match.

However, AI innovation is increasingly gaining global traction, and in the past year, competing models were developed in the Middle East, Latin America, and Southeast Asia. In addition, the past year saw international organizations including the Organisation for Economic Co-operation and Development (OECD), the European Union, the United Nations, and the African Union all release AI governance frameworks focused on transparency, trustworthiness, and responsibility.

The Open-Versus-Closed Debate

As the AI race speeds up, the debate between open versus closed models has garnered particular attention, with outcomes that will have direct impacts on AI development in the Global South.

Often perceived as existing along a spectrum, open AI models can be systems that publish the complete model architecture, just the source code, or just the data. These models can be downloaded by anyone to be modified or built upon (with some restrictions). Open foundation platforms such as Llama and Mistral allow users to see, modify, and apply the models for their own needs. Inversely, closed systems do not make their design public, and users only interact with the model’s interface. For example, one could ask OpenAI’s ChatGPT to draft a paper of a specified length on the impact that AI could have in developing countries. ChatGPT would write it, and although the user could see the sources from which information was gathered, they would not be able to see the code the model used nor the data with which that model was trained to come to the assumptions that undergird the paper.

▲ Figure 1: The Open and Closed AI Spectrum

In fall of 2024, the CSIS Project on Prosperity and Development hosted a private roundtable to discuss the comparative value of open and closed AI systems and their role in global development. During the conversation, advocates for open AI models highlighted their transparency, which they saw as a benefit and a catalyst for innovation. Open systems can allow for greater access by sidestepping the steep development costs associated with building computing capacity and initially training a model. Large language models (LLMs) are expensive to build but relatively cheap to fine-tune once established. Open models also allow a more diverse pool of developers to stress test a system, leading to fewer vulnerabilities and glitches over the long term.

However, critics of open AI models see the same transparency as a risk, arguing that the code could be used by malicious actors to create deepfakes and orchestrate more advanced cyberattacks, along with other harmful activities. Open approaches also present opportunities for adversaries to advance their own AI capabilities by co-opting technologies. Recent examples, such as the Chinese development of a potential military tool using Meta’s Llama model, prove that these concerns are not unfounded. However, proponents of open AI systems argue that closed systems are unlikely to be a barrier for many of the worst actors anyway—after all, thieves undeterred by locks on a door will still often break through a window. In the meantime, aspiring innovators with little access to the financing, technical skills, and infrastructure needed to develop their own models are likely to be disadvantaged.

Historically, open models have lagged behind closed models (in terms of both time and compute, or processing power) because developers have a financial incentive to keep their most advanced models private. While the timing delays remain significant, with the most recent studies citing that open systems continue to lag approximately 15 months behind closed systems, the differences in computing power have grown more negligible.

U.S.-China AI Competition

The United States has wielded export controls to try and limit the development of advanced AI systems by its adversaries. By withholding technologies with the best processing power, the United States could limit the ability of other actors, namely China, to perform on par with U.S. innovators.

The Biden administration took a hard-line approach in its attempt to limit China’s technological development. As one of the last acts of Biden’s presidency, his administration announced a sweeping set of guidelines, commonly known as the AI Diffusion Rule, dictating who can import advanced semiconductors under what conditions, and in what ways the technologies can be shared. President Trump, while keeping many of his predecessor’s export controls, struck down the AI Diffusion Rule two days before it was set to take effect in May 2025, claiming to be taking a more business-minded approach to the technology. He also designated David Sacks as the first “White House AI and crypto czar.” Additionally, as the One Big Beautiful Bill Act (H.R. 1) worked its way through Congress, the U.S. AI regulation debate was brought to center stage, with the House and Senate taking different stances on state-level domestic AI regulation. Ultimately, the language that became law did not include a provision initially passed by the House that would have specifically prohibited states from “limiting, restricting, or otherwise regulating” AI for 10 years.

In July 2025, President Trump announced a roadmap entitled America’s AI Action Plan, in addition to a series of executive orders, which structures the administration’s AI posture toward limited regulation. The plan, divided into three thematic pillars—(1) innovation, (2) infrastructure, and (3) international diplomacy and security—recognizes the geostrategic importance of maintaining U.S. AI dominance, encourages the creation of a “full-stack AI export package” for allies and partners, and promotes U.S. open AI models as the global standard for future AI development. These actions come at an incredibly dynamic time, in which the United States’ approach to AI regulation and development, both internationally and domestically, is in flux alongside the technology itself.

AI export controls, which began under Biden and have continued under Trump, are inherently porous, and adversaries can acquire greater processing power through any number of legal and illegal means. The creation and success of DeepSeek’s R1 model is a cautionary tale illustrating the challenges associated with effective export controls and an inspirational story for those looking to build their own models to rival the best U.S. ones. On par with the most advanced U.S. models from 2024, DeepSeek, developed in China, showed that the country was further along in the AI race than previously thought. In addition, China is investing substantially in building its own data centers, strengthening its power sector, and designing its own chip manufacturing capabilities—all in an effort to detangle itself from Western reliance. China has sought to capitalize on DeepSeek’s success to show that Beijing can be an alternative market for AI innovation and a strong tech partner for those looking to enhance their AI capabilities.

Despite its progress, China continues to rely on smuggled U.S. chips to power its technology. Only two of the twenty-two models developed exclusively by Chinese AI labs in 2025 were trained with Chinese chips. In 2024, hundreds of thousands of chips, totaling millions of dollars, were smuggled into China through a mix of shell companies, varying distributors, and mislabeling techniques. The Trump administration recently announced it will allow Nvidia to begin selling lower-level chips, known as H20, to China, a move likely to further underscore China’s dependence on Western AI technology.

Choosing which geopolitical AI “road” to take has already become a defining strategic decision for LMICs, one that will impact not just their own digital futures but also the United States’ and China’s soft-power influence across the Global South. China has worked to position itself as a global leader in AI cooperation through its cosponsorship of the AI Capacity-Building Action Plan for Good and for All, which was officially announced in September 2024 and emerged from the UN General Assembly resolution “Enhancing International Cooperation on Capacity-Building of Artificial Intelligence.” With this positioning, as well as its Digital Silk Road initiative—a comprehensive framework for exporting Chinese digital technologies that creates technological dependencies worldwide—China is asserting its technological influence globally. Simultaneously, U.S. export controls and the high costs associated with the adoption of Western AI capabilities have the potential to strengthen China’s attractiveness to LMICs. This is true even in the face of acute drawbacks associated with Chinese offerings, including technological dependence on Chinese systems and workers, as well as the potential for surveillance. LMIC governments are at a crossroads: They must choose between Western systems that are sometimes tortuous to deploy and Chinese alternatives that are potentially cost-effective, but carry long-term implications for technological diffusion, data sovereignty, and geopolitical alignment.

Developing Countries Weigh the Options

It is against the backdrop of U.S.-China competition that developing countries are making decisions about their AI futures. Beyond the technical choices countries have to make, the political consequences of picking one technological system over another can have long-lasting implications for the country’s geopolitical relationships. Countries working to build out their AI ecosystems are also forced to weigh the costs and benefits of prioritizing AI development versus financing other services such as economic assistance or paying down their debt.

Challenges Faced

AI development faces many challenges in LMICs, including high adoption costs, limited infrastructure and energy capabilities, low digital literacy rates, and a lack of coherent AI investment strategies. Costs associated with building the computing capabilities, data centers, supply chains, and technical knowledge needed to design AI models remain high, and are only increasing: Frontier AI model costs are estimated to have grown at a rate of two to three times per year for the past nine years. This is not to say that developers cannot build cheaper models, but to build a model on par with the most advanced systems today requires more and more capital as the models get more sophisticated.

The lack of foundational digital infrastructure and energy also poses significant challenges for widespread AI adoption in the Global South. AI requires stable internet access, access to digital technologies such as smartphones and computers, and consistent electricity, which is not universal. Across sub-Saharan Africa, for example, only one-quarter of the population has access to reliable internet, and there is a 29 percent gender gap in mobile phone usage. Energy consumption also poses a major challenge. Global estimates by the International Energy Agency (IEA) indicate that data center electricity consumption alone reached 415 terawatt-hours (TWh) in 2024, approximately 1.5 percent of electricity demand worldwide, and this figure is expected to triple by 2035. To put that in perspective, the total energy consumption of households in sub-Saharan Africa is expected to reach between 430 and 500 TWh by 2030. The enormous capital required to develop foundational AI infrastructure therefore presents a major barrier to entry for developing countries.

The domestic investment trade-offs and resource constraints LMICs face illustrate the opportunity cost associated with investing in AI systems. Although AI investments could possibly lead to an influx of foreign capital, future domestic resource mobilization, and more effective programs over time, the same limited resources could also be allocated to servicing onerous debt obligations or expanding domestic services more connected to immediate improvements in human welfare. Argentina provides an interesting example of this policy juxtaposition: While President Javier Milei has been a champion of Argentina becoming the region’s AI powerhouse, the country’s economic instability has led to significant cuts in public funding for education and the sciences while also leading to brain drain of its brightest engineers. These tensions raise the question of whether Argentina can realize its digital ambitions without undermining the very conditions it needs to support them.

The recently proposed Africa AI Fund offers an example of how governments might attempt to address these trade-offs through catalytic economic growth. The fund is designed to invest $60 billion in developing Africa’s AI ecosystem; in the process, it seeks to create 500,000 jobs annually and lift 11 million people out of poverty by 2030. However, even with significant capital from OECD nations and private philanthropic partners, LMICs would need access to significant investment capital to capture AI’s benefits, capital that could be otherwise allocated to debt repayments and other pressing domestic development needs.

The Role Open-Source AI Could Play

In addition to the cost of adopting foreign AI models, depending on such models poses major risks to data sovereignty, economic development, and regulation. These risks are behind a recent push for sovereign AI ecosystems, seen as a new horizon in self-determination. Sovereign ecosystems are defined loosely as a country’s efforts to develop, control, and govern its own AI capabilities without depending entirely on foreign support. This can range from building data centers and designing governing frameworks to training AI models in local dialects with local references.

As LMICs aspire to build out the infrastructure, energy, and enabling environment needed for AI implementation, open AI systems offer key opportunities, particularly as many of these countries lack the funding to build their own models. Instead of prioritizing the development of completely new technologies, building country- or region-specific facilities to train open models to fit specific local needs could reduce costs. Using already developed LLMs, while still developing the necessary infrastructure and programming talent for localization, can leapfrog the time, expense, and technical and infrastructural capabilities required to create models anew. Similarly, small language models, which are trained on fewer parameters than LLMs, are estimated to consume less than 20 percent of the energy required to run a LLM, and have the potential of limiting costs while providing a more curated, culturally relevant experience for AI users in languages or communities that are often overlooked in traditional cases of technology transfer.

While LLMs lend themselves to the development of educational tools, given their prevalence in research and writing today, computer vision (CV) models offer another avenue for AI penetration in developing countries. Meta’s DINOv2 and Segment Anything, and Amazon Web Services’ SageMaker are key examples of open-source CV models. Unlike LLMs, CVs train machines to distinguish and derive information—the distance of an object, whether it is in motion, or if it has a defect or issue—from visual inputs such as photos or videos. CVs can be applied to agriculture and disaster response. Through machine learning (when the technology teaches itself to identify a specific picture) and convolutional neural networks (a process that allows the machine learning model to break down a picture to its individual pixels), CVs can be trained to recognize and predict crop failure, weather events, and damages to infrastructure and land, enhancing agricultural production and streamlining emergency relief. While these new tools are advancing geospatial capabilities, limited datasets continue to be a barrier to predict and improve agricultural and disaster modeling.

Potential and Progress

Innovators across the Global South are applying AI solutions to local problems—with open-source models offering advantages in adapting to local cultures and languages while preventing vendor lock-in. In Africa, AI hubs are popping up to address development issues while streamlining productivity, reducing costs, and improving access to information. AI has particular applications in the agriculture, healthcare, and education sectors—three areas that face significant challenges across Africa.

Agriculture, which accounts for 42–48 percent of Africa’s workforce and averages 17 percent of its GDP, intersects with several challenges, including limited access to resources, unpredictable weather patterns, and low productivity. Those issues are exacerbated by climate change: In 2022, African countries incurred close to $9 billion in losses and damages due to climate-related events. AI-powered solutions in areas such as crop monitoring, weather forecasting, and pest management offer the potential to optimize agricultural practices, improve yields, and build resilience against mounting climate impacts.

In addition to its application in the agriculture sector, AI can help to bridge the healthcare and teacher gaps in developing countries. By 2030, the UN Educational, Scientific and Cultural Organization estimates that an additional 44 million teachers will be required to meet primary and secondary school demand, a third of which will be needed in sub-Saharan Africa; in the same timeframe, the global health sector is expected to face a workforce gap of up to 10 million healthcare professionals. Applying AI solutions to telemedicine, disease detection, and health monitoring can increase patients’ access to care. Educational tools built on open AI models, such as specified learning plans and adaptable chatbots that can answer questions relating to a student’s curriculum within their linguistic and cultural context, are also helping in the face of educator shortages across the Global South.

In tandem with the specific AI applications being developed, governments and international organizations are working to forge regulatory frameworks to protect people’s data, encourage investment, and build more inclusivity across communities. Twelve new AI strategies were published at the country level in 2024, more than half from LMICs, according to the 2024 Government AI Readiness Index. These strategies focus on building open-source cooperation across sectors to share resources, technical assistance, and best practices, illustrating a growing AI governance trend gaining traction across the Global South. For example, the African Union’s Continental Artificial Intelligence Strategy calls for a unified regional approach to “strengthen regional and global cooperation and position Africa as a leader in inclusive and responsible AI development.” The strategy places an emphasis on promoting national and regional data pools and underscores the need for democratized datasets to facilitate innovation and economic development.

Country Spotlights: Kenya and Nigeria

Kenya and Nigeria are becoming hubs of AI innovation geared toward agriculture, healthcare, and education. Through a mix of public-private partnerships, local innovation hubs, and the adoption of open-source AI models, Kenyan and Nigerian entrepreneurs are building AI solutions to address local challenges, as outlined in Table 1.

▲ Table 1: AI Innovation Across Industries in Kenya and Nigeria

Both governments have recently launched AI knowledge centers to improve public awareness, digital literacy, and collaboration. The Nairobi-based Centre of Competence for Digital and Artificial Intelligence Skilling, established in partnership with the government of Kenya, the UN Development Program, and Microsoft, works to drive digital inclusion by building public-private partnerships and strengthening local technology ecosystems. Kenya’s strong regulatory and legal framework, in addition to its substantial digital infrastructure investments, has made it a digital leader on the African continent. Similarly, the government of Nigeria, with support from the Gates Foundation, launched the Nigeria AI Scaling Hub in June 2025 to accelerate AI adoption in key areas including agriculture, healthcare, and education.

Advancing AI Capabilities in the Global South Through Private Sector Partnerships

LMICs are partnering with U.S. tech firms to jump-start their sovereign AI ambitions. By pairing foreign capital and expertise with domestic oversight, developing country governments aim to have more autonomy over regulations, with local economies capturing a greater share of AI’s vast social and financial upside. In infrastructure, U.S. cloud providers are anchoring regional computing hubs through billion-dollar investments into data-center capacity. These projects have catalyzed local economic growth and seeded additional expansion: AWS Africa’s Cape Town cloud region, opened in 2020, has already added approximately $673 million to South Africa’s GDP and enabled low-latency access to millions of end users in the region; Google’s new Johannesburg cloud region has been operational as of early 2024; and Microsoft and Abu Dhabi–based G42 are investing $1 billion in a geothermal-powered campus in Kenya. OpenAI’s newly launched OpenAI for Countries initiative, a subproject within its $500 billion Stargate Project that was announced by President Trump, catalyzes the development of safer, more scalable national AI ecosystems through infrastructure collaboration and partnership with the U.S. government. The project is aimed at helping build data center capacity, providing AI in the context of local languages and cultures, securing models both physically and by increasing democratic input, and developing national AI ecosystems and jobs.

Against this backdrop, there is also a significant need for investment into human-capital pipelines and access to open-source backends to develop and train language models on localized data for localized uses. Microsoft has pledged to support one million Nigerians in AI upskilling by 2027, while Google has committed multimillion-dollar investments for similar purposes. The development of open-source platforms like Hugging Face’s model hub or Meta’s Llama-3 will allow this technology to be deployed and iterated upon in ways that encourage entrepreneurs and LMIC governments to better solve their challenges and improve quality of life for their populations.

Recommendations to Strengthen U.S.-LMIC Tech Partnerships

As the United States looks to strengthen its influence and build partnerships with developing countries, technology will play a large role. By wielding its status as a leader in AI development, and drawing on the financing, R&D, and influence of its private sector, the United States can help developing countries access the keys to unlock their digital transformation. Additionally, the United States can lead by example, strengthening its own AI enabling environment by bolstering intellectual property rights and creating competitive markets that drive innovation. As the United States works to help developing country partners build out their own AI ecosystems, policymakers should consider the following recommendations.

1. Promote good governance and regulation.

AI governance frameworks that prioritize transparency, democracy, and the ethical use of technologies not only protect individual rights but also support an enabling environment that encourages foreign investment in local AI development. However, capabilities for supporting AI-related partnerships and governance reforms have been stalled or lost as a result of the dismantling of the U.S. Agency for International Development (USAID). These capabilities should quickly be reconstituted.

The United States’ presence internationally has changed significantly in recent months. The AI export stack outlined in the new AI Action Plan bundles a comprehensive set of AI technologies, infrastructure, and expertise, and marks a step in the right direction to ensure the United States remains an AI leader on the global stage. However, the plan currently lacks a developing country angle—specifically, it does not address the enabling environment needed for the AI stack to succeed. As such, the U.S. Department of State, together with the U.S. Trade and Development Agency, the U.S. Development Finance Corporation, and the U.S. private sector, should work with developing country partners to develop regulatory frameworks that ensure the safety of individual users, align with democratic values, and encourage future U.S. private sector investment. With the Office of the Special Envoy for Critical and Emerging Technologies slated for elimination under the current State Department reorganization proposal, the realigned Bureau of Cyberspace and Digital Policy could lead this work in close partnership with the “F family” of bureaus managed by the proposed under secretary for foreign assistance, humanitarian affairs, and religious freedom. By combining the assistance provision expertise housed in these bureaus with the technology expertise of the Bureau for Cyberspace and Digital policy, the United States will be better positioned to assist developing countries in their technological journeys and facilitate the export of more comprehensive AI stacks by the United States.

2. Support the sharing of U.S. private sector technical expertise.

Because of the talent, financing, and entrepreneurship that the United States has historically attracted, it is leading the AI race. The U.S. tech sector’s expertise makes the United States a desirable trading partner for developing countries. The AI Alliance, jointly founded by Meta and IBM, is an example of how the U.S. private sector can play a role in informally shaping international AI norms while encouraging innovation. The alliance, consisting of over 140 members from 23 countries, aims to bring open-source developers into dialogue with each other, and with researchers and organizations around the world, to foster greater transparency, safety, and security in AI design. By sharing technology, expertise, and lessons learned through similar public initiatives, public-private partnerships, educational exchanges, or direct investment, the United States and its private sector can help countries harness the power of AI while promoting principles of transparency and safety and strengthening its relationships across the Global South during an era of intensifying geostrategic competition.

3. Encourage open-source AI model development and application.

While they carry real risks relating to security, usage, and competition, open AI systems can foster innovation and lower the barrier to entry for countries and companies that would otherwise not have the skills, financing, or support to build out their own models. To combat the potential downsides, AI governance frameworks need to address the risks associated with open-source models while building an environment that supports the sharing of technologies for innovation. The U.S. National Institute of Standards and Technology (NIST) has published an AI Risk Management Framework to help organizations identify the different risks AI can pose through its entire lifecycle, ranging from setting benchmarks for transparent open-source model development to the distribution of the technology for public use. Promoting NIST-aligned risk management frameworks internationally could help establish a baseline for trust and safety, enabling more responsible AI diffusion worldwide. To further encourage the adoption of U.S. models, the United States should support international collaboration through active participation in multilateral forums where countries are cocreating governance best practices. These efforts would reinforce global trends toward shared frameworks and cooperation on governance, such as the African Union’s Continental Artificial Intelligence Strategy.

4. Mobilize U.S. private sector investment.

By working with the U.S. private sector to invest in AI innovation and infrastructure projects in the Global South, countries are able to further develop their economies, increasing commercial connections to U.S. markets and related economic growth while, over time, strategically reducing the need for continued U.S. economic development aid. Public-private partnerships and initiatives such as the U.S.-India Roadmap on Accelerating AI Infrastructure, announced during Prime Minister Narendra Modi’s February 2025 visit to the United States, are adding millions of dollars to local economies, strengthening workforce capabilities and digital skills, and developing the infrastructure needed for countries to succeed in the dawning era of AI. The U.S.-India roadmap, part of the larger U.S.-India Transforming the Relationship Utilizing Strategic Technology (TRUST) initiative, aims to enable investment in U.S. AI technologies in India through the development of data centers, increased access to computing power, and lowered regulatory barriers. In order to further mobilize such private sector investment in developing economies, the United States—through investment incentives from the U.S. International Development Finance Corporation and technical expertise shared by the U.S. Trade and Development Agency, the Millennium Challenge Corporation, and the Departments of State and Commerce—should adopt strategies similar to the U.S.-India roadmap that encourage innovation and the tailored sharing of technology. Such efforts across developing countries in particular are made more difficult by the dismantling of USAID, which had painstakingly developed relevant expertise over the past fifteen years—engaging with countries on their digital transformations while collaborating with Silicon Valley firms on scaling applications of innovative technology. As is the case with technical support for governance reforms, the dismissal of relevant expertise that existed within USAID to support private sector engagement in developing countries has led to a loss of U.S. government capabilities that should be restored and further strengthened.

Conclusion

Developing robust, inclusive, and democratic AI ecosystems around the world will ensure that the greatest number of people will be able to benefit from the technology. While competing challenges continue to hinder the ability of developing countries to integrate AI into daily life and develop solutions that are tailored for their specific challenges, such countries risk falling further and further behind. AI has a wide range of applications for pressing international development challenges in many sectors, including healthcare, education, financial inclusion, manufacturing, energy, disaster risk management, agriculture, and food security. As such, it is imperative that the United States and other developed economies reframe their tech relationship with developing countries to one of collaboration and codevelopment.

Noam Unger is vice president of the Global Development Department and a senior fellow with the Project on Prosperity and Development at CSIS.

Madeleine McLean is a program manager and research associate with the Sustainable Development and Resilience Initiative and the Project on Prosperity and Development at CSIS.