Policies for Artificial Intelligence (AI).

AI technology has been theorized and developed since the middle of the last century, but has only recently entered everyday life and the policy realm. In 2017, Canada became the first country to officially issue a national AI strategy. Since then, AI has attracted significant attention from policymakers, with at least 1,900 new policy instruments, and 89 national strategies. Despite this rapid rise, AI policy is still a relatively new field of action, with profound uncertainties about what is needed and what works and what does not. With the integration of AI into an increasing number of activities. Governments need to respond as a matter of both public concern and economic development. Increasing public awareness and concern about issues such as labour protection, human rights, unethical use, personal autonomy, data privacy and bias and discrimination have amplified attention paid to AI. While uncertainty and risks of failure are significant, inaction could result in even greater costs. Traditional policy and regulatory models struggle to match the speed, autonomy and opacity of AI systems, posing challenges for Governments, businesses and the international community. Policies for frontier technologies and AI need to be flexible and regularly updated. To date, most AI policies have been produced by developed countries. At the end of 2023, about two thirds of developed countries had a national AI strategy. Only 6 of the 89 national AI strategies were from LDCs (figure IV.5). Bangladesh and Sierra Leone took the lead in 2019 and were joined by four other LDCs in 2023, an uptick that may signal the beginning of greater LDCs participation in AI policy-making discourse, although these six countries form only around one eighth of LDCs. LDCs and developing countries need to move quickly to align AI adoption and development with their national development goals and agendas. Following the path set by others may not fulfil their needs and priorities.

Figure IV.6 shows the most common policy instruments. More than one third are related to national strategies and agendas, AI-related regulations or public consultations. This includes gathering information on technological trajectories, addressing social concerns and anticipating possible opportunities and downsides. Although around one third of developing countries have strategies and plans, these may not go beyond the declarative stage if they are not complemented by sufficient resources and instruments for implementation. Policy instruments also support earlystage science and technology efforts, including networking and collaboration, public awareness campaigns and outreach activities to engage civil society. It is important to connect diverse actors in the AI innovation ecosystem, enabling idea exchanges, resource-sharing and collaboration, in order to identify gaps, promote best practices, prevent duplication and ensure efficient resource use.

 To support the development and diffusion of AI, developed countries are more likely to use financial instruments, such as competitive grants for public research and for business R&D and innovation, as well as student fellowships, along with policies to support the development and uptake of AI through computing and research infrastructures. A greater proportion of instruments directlyfunding STI and AI infrastructure can be related to the larger budgets dedicated to R&D in developed countries. In contrast, developing countries are more likely to target the use of AI in the public sector. Incorporating AI into e-government practices can expedite government processes, help overcome limited resources or bureaucratic backlogs and help learn about AI through its use. However, this should not be at the cost of direct and practical interventions to support STI related to AI and create a supportive environment for business innovation that turns declarations into reality.

The rise of digital technologies has made timely information and research results more easily accessible, helping diffuse new ideas and enabling a more participatory approach. In figure IV.6, this is reflected in the number of instruments targeting networking and collaborative platforms or public awareness campaigns to reach civil society. These platforms can also help address gaps in the AI ecosystem, helping to share best practices and reduce the duplication of efforts. Typically, the countries more prepared for AI governance are developed countries with higher per capita GDP (figure IV.7).

Readiness rises with GDP per capita and less advanced countries are in general unprepared to capitalize on AI opportunities and deal with risks, leaving them exposed to technological paths and regulations set by others. However, some countries at the same levels of income are achieving more. For example, Rwanda, which issued a national AI strategy in 2023, has a much higher AI governance score than other countries with similar GDP per capita. Other “overperforming” developing countries include Brazil, China, India and Singapore, which have policies and strategies that could offer useful lessons for other countries. 

Adopting – Policies targeting AI adoption should support the uptake and diffusion of AI products and solutions in the economy and provide upskilling and reskilling training to the workforce exposed to AI. By upgrading existing activities or enabling new ones, the diffusion of AI could move an economy towards the technological frontier.

Many developing countries, however, are still in the policy design phase, partly because they lack AI ecosystems that can provide the necessary expertise on bottlenecks, opportunities and the measures that favour AI uptake. While developing countries may prefer to initially grasp only the low-hanging fruit of AI adoption, this could limit their capacity to catch up. In the long term, their opportunities for learning through imitation are likely to be hindered by the rapid evolution of technology. Developing – Policies targeting AI development should expand the capabilities required to generate new knowledge, and create new prototypes, systems and applications. These could include networking and distributing computing power across a country. Developed countries have done so in order to keep pushing the technological frontiers. The two approaches are not, however, mutually exclusive and countries need to strike a balance between them. Developing countries may find it less challenging to support adoption by responding to particular sectoral needs, while taking a targeted approach to trigger positive dynamics and improved innovative capabilities. Yet they also need to make long-term strategic plans to support AI development; otherwise, as latecomers, they may end up with few options.