Case studies of AI-related policies.

 This section discusses overarching approaches and strategies of the three main global markets: China, the European Union and the United States, then presents instruments that address bottlenecks at the three leverage points of infrastructure, data and skills (table IV.2).

For the digital economy, there are three main regulatory approaches. One option, as favoured in China, is direct intervention in support of national political goals using strict regulations. A second, as in the European Union, is strong regulations aimed at protecting fundamental rights and values. A third approach, favoured in the United States, involves a light regulatory framework. Recently, the development of AI and its wide-ranging societal and economic effects have influenced country strategies, with emerging similarities in approaches. The first step of a national AI strategy is to identify and address coordination failures and weaknesses in the innovation system. Governments can, for example, support applied research through project grants for AI-related business activities. Pilot AI use cases in particular sectors and knowledge and technology transfer mechanisms can contribute to accelerate the adoption of AI. Countries can consider a multistep approach, as in China, first incentivizing the private sector to adopt, adapt and develop AI, and subsequently supervising and regulating the AI industry. Governments need to promote good practices and enforce rules and standards, while revising regulations and policies to adapt to changing circumstances. For example, the European Union provides a coherent framework integrating new legislation as it emerges, to address issues such as consumer protection, and regulating platforms to counterbalance concentration and ensure data protection. Policy formulation and implementation are interactive and iterative processes that require continuous evaluation, and expectations need to be aligned with feasibility.  

Failures should be accepted, as they are with regard to new ventures in the private sector, but evaluation mechanisms should be put in place to improve design and implementation. Currently, only about 10 per cent of the AI policies surveyed by OECD have been evaluated, based on data from the AI Policy Observatory.

The Government of China has taken an increasingly active role in AI. In 2017, it set out a long-term strategic plan to transform China by 2030 from an AI contributor to a primary AI innovator.

The plan is: • Technology-led – deploying forward-looking R&D in key frontier domains and achieving transformational and disruptive breakthroughs. • Systemic – formulating targeted strategies for different technologies and industries. • Market-oriented – fostering commercialization of AI and creating competitive advantages in related technologies. • Open – advocating open-source approaches to enable industry, academia and research collaborations.

 China is now formulating industry standards and expanding regulatory oversight, and has recently moved to a more direct supervision of AI, introducing some of the world’s first binding national regulations, defining requirements for how algorithms are built and deployed and establishing the information that developers must disclose to the Government and the public. In 2023, the Cyberspace Administration introduced Interim Measures for the Administration of Generative Artificial Intelligence Services, for regulating research, development and the use of GenAI (Cyberspace Administration of China).  The measures impose various obligations on GenAI providers to ensure that models, contents and services comply with national requirements and uphold “core socialist values” and national security. They also aim to ensure the transparency of GenAI services and the accuracy and reliability of generated content, to prevent discrimination and respect intellectual property and individual rights. In this last aspect, the measures echo earlier provisions targeting deepfakes and fake news. In 2024, the Government launched a National Data Bureau to coordinate and support the development of foundational data systems, and to integrate, share, develop and apply data resources. China relies on a series of technical and administrative tools, such as disclosure requirements, model auditing mechanisms and technical performance standards, as well as measures to ensure that public bodies are responsive to technological development. Focusing on particular emerging issues and technologies reduces the burden of generalization but demands a high level of responsiveness to technological advances and strong coordination among public bodies.

In 2024, the European Union passed the AI Act, which defines rules according to the associated level of risk, namely, unacceptable, high, limited or minimal (European Parliament and Council of the European Union, 2024; O’Shaughnessy and Sheehan, 2023). Most applications, such as video games or spam filters, fall in the minimal risk category, and companies are only advised to adopt voluntary codes of conduct. The Act allows high-risk AI systems but says that these should include complete, clear and accessible instructions, which should be stored in an open database maintained by the European Commission in collaboration with member states. The Act bans uses that present unacceptable risks, such as cognitive behavioural manipulation, social scoring, biometric identification and categorization, as well as remote biometric identification systems such as facial recognition. This is known as a risk-based approach. The AI Act builds on previous legislation such as the General Data Protection Regulation of 2016, which guarantees privacy and respect for human rights (European Parliament and Council of the European Union, 2016). The Digital Service Act of 2022 is aimed at establishing a level playing field, to promote innovation and competitiveness in information services, from websites to digital platforms, and stop large providers from imposing unfair conditions that damage other businesses or limit consumer choice. 

The European Union has also revised its industrial strategy to address external dependences on critical technologies. Strategic areas related to the AI value chain are critical raw materials, semiconductors, quantum technologies and cloud computing. In these areas, the European Union is building industrial, research and trade policies, fostering co-investment across member states and bringing together stakeholders in industrial alliances. In 2023, to strengthen competitiveness and resilience in semiconductor technologies and applications, the European Union passed the European Chips Act, aiming to mobilize more than €43 billion of public and private investments and setting out measures to prepare for, anticipate and respond to possible supply chain disruptions, while strengthening its technological leadership. The European Union has also allocated funds for AI research and innovation. The European Research Executive Agency manages more than 1,000 research projects, with pioneering projects in AI and quantum technologies.

  

In 2022, the United States Congress passed the CHIPS [Creating Helpful Incentives to Produce Semiconductors] and Science Act to boost scientific research and advanced semiconductor manufacturing capacity. The act was motivated by increasing dependency in chips manufacturing and the fact that federal R&D spending had neared its lowest point in 60 years, and targets frontier technologies, including AI. Of the $250 billion budgeted, 80 per cent are allocated to research activities and the rest to tax credits for chip manufacturers. The Act exemplifies key aspects of policies for emerging technologies. It adopts an anticipatory approach, supporting technologies that could shape future industries. It addresses coordination failures, and leverages complementarities through a supply chain approach, supporting activities from hardware production to computing infrastructure, research, and skill development. New talent will be trained through a national network for microelectronics education, as well as cybersecurity workforce development programmes. 

To retain talent, an AI scholarship programme has been set up for students who committed to a period of government service. The Act also promotes safe and trustworthy AI systems and the collection of best practices for artificial intelligence and data science. Finally, it envisages public–private partnerships that would establish virtual testbeds to examine potential vulnerabilities to failure, malfunction or cyberattack. The Blueprint for an AI Bill of Rights noted that AI and automated decision systems should not advance at the cost of civil rights, democratic values or foundational American principles, and set out principles to guide the design, use and deployment of automated systems to protect the public. Action is also being taken by individual states. In California, for example, an AI bill in 2024, required firms to commit to model testing and the disclosure of safety protocols and made compulsory a series of requirements that were previously only voluntary. This could represent a major shift in the way emerging and potentially disruptive technologies are dealt with in the United States. Figure IV. summarizes the main elements of AI policies deployed by China, the European Union and the United States. All are taking a cautious approach to regulating AI development, alongside substantial public investments across the AI supply chain, from semiconductors to data centres, and in research and development, to foster the emergence of new industries. Moreover, they aim for the inclusive integration of AI into both economies and societies, to benefit a wide range of stakeholders. These commonalities highlight key elements to consider in both national and global AI policy strategies. 

AI policies in major economies can create significant spillover effects, shaping the policy choices of other countries. As leading countries set higher benchmarks, particularly in boosting competition and prioritizing R&D, not all countries are equally positioned to keep up. Many may struggle to match increasing R&D budgets, and the focus on future technologies can deepen disparities, widening the gaps between advanced economies and those working to catch up. This highlights the challenges faced by smaller or less advanced countries in keeping pace with global innovation leaders.

AI infrastructure can be classified under the two board category of digital technology and computing power.

Brazil – In 2023, the New Growth Acceleration Programme planned a $5.7 billion investment to foster the transition to a digital economy through public–private partnerships for digital infrastructure; the federal Government would contribute about 44 per cent of the overall budget, State owned companies, 20 per cent, and private companies, 36 per cent. The plan is to expand 4G networks across the country, deploy new 5G networks and reinforce infrastructure with fibre-optic cables, such as the 587 km-long cables that will connect the capitals of two northern states, Amapá and Paraná, on opposite sides of the Amazon delta. This connectivity upgrade is aimed at reaching all public schools and healthcare units, contributing to the modernization of the public sector (Brazil, Federal Government, 2024).

 Côte d’Ivoire – Targeted infrastructure can support the adoption of AI in particular sectors. For example, the e-Agriculture project is aimed at increasing the use of digital technologies and improving farm productivity and access to markets. This is being pursued by improving Internet coverage and adoption, fostering the use of large-scale digital platforms, rehabilitating rural access roads and adopting sustainable digital services to diffuse e-agriculture. Focusing on both physical infrastructure and digital services, the project represents a value-chain approach that can respond to community needs (World Bank, 2024).

 Japan – The High Performance Computing Infrastructure project strengthens national computing capacity for AI development. The project uses an existing supercomputer and connects major universities and national laboratories via a high-speed network (Research Organization for Information Science and Technology, 2024). By decentralizing access and networking institutions the project increases computing power availability and supports innovation in computing-intense sectors, increasing the number of new actors in the AI ecosystem. Decentralized organizational systems and distributed networks are crucial aspects of the digital revolution and a cornerstone of advanced AI ecosystems.

Republic of Korea – The K-Chips Act increases tax credits for investments in semiconductor enterprises and other national strategic technologies, with a focus on SMEs (Pan, 2023). The policy supports the development and production of essential hardware components of the AI value chain by streamlining regulation and standardization in the field of microchips, to provide a common and clear playing field for business development. 

Data is a key production factor in the knowledge economy. Many countries already had data policies in place before the advent of AI, but will need to update them, while others still lack national data frameworks. Data policies should ensure that databases are interoperable and available across the economy, with privacy protection for both inputs and outputs, relying on consent and taking account of possible biases. AI systems add concerns related to ownership, while also raising questions of intellectual property or fairness and accountability when generating data and decisions. Supporting AI development may require rethinking intellectual property provisions and creating mechanisms to facilitate public–private collaboration. Such efforts should promote AI innovation while safeguarding human rights and addressing potential vulnerabilities and malfunctions. Policies should also respond to the international and transboundary nature of AI. 

Using cloud computing available from international markets can reduce costs, but it is important to avoid increasing data and information dependency and stifling the future development of a domestic service market. Open data refers to data that is openly accessible, exploitable, editable and shared by anyone for any purpose. An open-data hub integrates disparate data into a single new system homogenizing data and thereby guaranteeing compatibility, to allow for real-time processing from different entry points. A hub can also integrate tools with which to process data or develop applications; for example, the GitHub open data hub provides open-source AI tools for running large and distributed AI workloads. Countries need to consider all levels of the data value chain. Policies should clearly define which types of data can be made publicly available, and how they should be handled, and favour standards for data and metadata. Countries can also collect and provide open data, either through AI-specific programmes or through open-data initiatives and hubs, to streamline data integration, storage, access and collaboration. This could improve transparency, promote innovation and encourage public engagement in the adoption and development of AI. Governments can also rely on industrial players to leverage existing strengths by supporting platforms for data exchange and aggregation and for data monetization and the development of AI for particular uses. Different types of data have their own requirements. In particular, for data on humans, or AI applications making decisions for humans, there should be higher standards for privacy and responsibility, and accountability in case of errors. Policies and standards can be developed through public consultations and open forums, to incorporate the views and concerns of different stakeholders, increase accountability and transparency and foster trust (table IV.4). Data can have broad social value because they are non-rival, namely, the use of a data set does not preclude its availability for other uses. However, the strong market power of large digital corporations may limit the capacity of developing countries to maximize benefits. UNCTAD, in a recent study, analysed the relationships between data and sustainable development.

Chile – The Ministry of Science, Technology, Knowledge and Innovation, and the Ministry of Economy, Development and Tourism have set up the Data Observatory, a public–private– academia collaboration that seeks to maximize the benefits from data for science, research and productive development. As a multi-stakeholder organization, the Observatory leverages the competences and resources of a variety of actors for developing STI and data-based services and analyses in different fields, from natural science to urban planning. It uses open-data platforms that facilitate the participation of small providers and supports projects and initiatives related to data analysis for social impact.

Germany – The Federal Ministry of Digital Affairs and Transport has launched Mobility Data Space, which brings together automobile companies, organizations and institutions that wish to monetize their data, seek data exchanges that bring mutual benefits or need data for innovative AI mobility solutions. A market-based platform, it incentivizes participation by offering the potential for financial remuneration – representing a model that leverages existing industrial strengths to support the diffusion of AI. 

India – The Council of Medical Research has issued Ethical Guidelines for Application of Artificial Intelligence in Biomedical Research and Healthcare, to direct AI adoption and development involving humans or their data (INDIAai, 2023). These recognize the importance of processes for safety and minimizing risk to prevent unintended or deliberate misuses that can harm patients. Data sets used by AI should avoid biases by adequately representing the population and guaranteeing the highest privacy and security standards for patient data.

Colombia – The Data Protection Authority has created a Sandbox on Privacy by Design and by Default in Artificial Intelligence Projects (Ibero-American Data Protection Network, 2021). This is an experimental space where AI companies can collaborate on solutions that respect personal information and rights, by design and in compliance with national data-processing regulations. The Authority accompanies the process and gathers information about possible regulatory adaptations, to keep pace with technological advances, thereby also making the sandbox a tool for policy learning. 

Singapore – In the Copyright Act 2021, Singapore redesigned the copyright regime to take account of how copyrighted works are created, distributed, accessed and used (Singapore, The law revision commission, 2021). The Act is aimed at making available large and diverse data sets for algorithmic training. The Act introduces an exception to the current regime that permits the copying of copyrighted works for the purpose of computational data analysis such as text and data mining and the training of machine-learning algorithms. It also introduces conditions and safeguards to protect the commercial interests of copyright owners.

AI has the potential to transform many industries in the near future, reshaping labour markets, altering tasks and changing required skill sets. Demand is increasing for skilled workers who can adopt and develop AI, including technical expertise in data science and AI skills for particular business operations. Countries need population-wide digital literacy, to ensure that everyone can take advantage of AI for work and personal life, and to have highly trained individuals who can develop AI systems and adapt them to particular needs. 

This should start with the inclusion of STEM and AI subjects at multiple levels within the national education system, from early education to adult learning. Introducing foundational data science and AI-related subjects in the early phases of education can help develop technology-savvy generations ready for AI-based businesses. Governments can also introduce or encourage programmes for retraining upskilled or displaced workers, with particular attention paid to women, who are underrepresented in both STEM and AI, and to older workers with low levels of digital skills, who are less likely to engage in such training. Policymakers can address concerns about diversity and inclusivity by empowering all demographic groups with the necessary skill sets to benefit or contribute to AI. By partnering with private institutions, Governments can also target particular sectors or industries. 

Philippines – In 2023, the National Economic and Development Authority published the Digital Workforce Competitiveness Act. The legislation puts human development at the forefront, aiming for equitable access and the provision of digital skills and competences that meet global quality standards to accelerate innovation and entrepreneurship. The Act targets particular digital skills, such as data analytics and AI or engineering and cloud computing, through upskilling, reskilling and training programmes, offering a variety of incentives to foster digital careers. The Act takes an anticipatory approach, envisaging the mapping of digital skills and technologies as the basis for formulating a road map that considers the evolution of jobs and skills. It also establishes an inter-agency council, including different state departments and agencies, which raises awareness about digital upskilling opportunities and coordinates actions, leverages complementarities, rationalizes policy interventions and provides a single-entry point for training, certification and scholarships. 

Spain – The National Plan for Digital Skills provides a list of actions and objectives to address gender bias in digital technologies (Spain, Ministry for Economic Affairs and Digital Transformation, 2021) and to increase the readiness of girls and women for AI To direct girls toward these disciplines, it introduces STEM subjects in primary education and includes programmes aimed at orienting women towards digital professions. The plan involves an analysis of the strengths and weaknesses of, opportunities for and threats to women’s participation in digital and technology careers.

 Ghana – To enable the younger generation to keep pace with a continuously evolving field, the Government has introduced coding and programming to the national education system and begun to train educators in how to teach them (Ghana, Ministry of Education). Moreover, subjects go beyond coding skills, to cover the fundamentals of how AI works, and concepts related to human, animal, robot and artificial intelligences, as well as weak and strong AI. The programme is gender responsive and is aligned with other initiatives such as the Girls-in-ICT programme (Ghana, Ministry of Communication, Digital Technology and Innovations, 2024), which has provisions similar to the National Plan for Digital Skills in Spain.

Nigeria – To foster the development of the AI ecosystem, the Federal Ministry of Communications, Innovation and Digital Economy launched the Nigeria Artificial Intelligence Research Scheme, aimed at providing financial support and facilitating knowledge-sharing and collaboration among individuals and organizations, to nurture new actors in the AI industry Nigeria launched the 3 Million Technical Talent programme to fund the training of selected fellows in 12 technical skills. The first phase of the programme is aimed at training 30,000 students and will then be scaled up. (Nigeria, National Information Technology Development Agency, 2024). The scheme provides scholarships to develop skills related to the digital economy, such as data science, AI and cloud computing. By fostering partnerships between high-skill AI researchers and businesses, the scheme is part of a broader strategy to build the workforce of the future.

The resurgence of industrial and STI policies, coupled with the rapid advancement of AI, has placed AI policies at the forefront of policymaking. AI policies are crucial in driving structural transformation, boosting productivity and tackling social, ethical and environmental challenges. As the global economy transits towards services and digitalization, Governments should adapt industrial and STI policies, to support the adoption and development of new technologies, as well as the dissemination and absorption of knowledge. Adapting to changing global conditions and harnessing frontier technologies requires swift and purpose-driven policy interventions. However, setting AI policies is not easy. When Governments need to provide public goods for these technologies, they have broad decisionmaking authority, but this is tempered by uncertainty regarding the trajectories and outcomes of policy decisions. Nevertheless, an anticipatory approach can help avoid the need to make corrections after most opportunities have passed. The unique characteristics of datadriven AI highlight the need for policy changes, with robust data governance, including regulations and standards for data-sharing and privacy protection. Additionally, the ability of AI to generate new data and concerns about deepfakes and misinformation require frameworks that regulate AI not only as a product but also within decision-making processes, ensuring transparency, explainability, ethics and accountability. However, considering the high level of concentration of AI markets, enforcement and regulation can be challenging for smaller economies. In this respect, later we will discuss AI policy efforts at the international level, offering suggestions of how the international community can support inclusive AI development that benefits all. AI is a pervasive technology that requires a whole-of-government approach, to align AI strategies with policies across sectors, including industry, education, infrastructure and trade. Doing so requires enhanced coordination, to leverage synergies among action plans. AI policies should go beyond incentives such as tax deductions, and incorporate regulation, governance and enforcement, to direct technological change and provide collective solutions to the major challenges of this century. Collaboration among stakeholders is essential to maximize societal benefits. To ensure effective adoption and development, successful AI strategies should also focus on the key leverage points of infrastructure, data and skills.