2. LAC Artificial Intelligence strategies

Regional collaboration for collective approaches

Governments around the world are adapting to the new possibilities offered by AI to transform government. As part of this process many governments have adopted international – often regional – strategies or other commitments for AI. For instance, in 2018, all European Union (EU) member countries signed the Declaration of Cooperation on Artificial Intelligence,1 committing them to work together to boost European AI capacity and adoption, ensure an adequate legal and ethical framework, and make AI available and beneficial to public administrations. The subsequent EU Coordinated Plan on Artificial Intelligence2 built on the declaration and seeks to “maximise the impact of investments at EU and national levels [and] encourage synergies and cooperation across the EU”. A 2021 review of the plan3 provided a set of joint actions for the European Commission and member states developed to create EU global leadership on trustworthy AI. In another regional effort, ten governments4 signed the Declaration on Artificial Intelligence in the Nordic-Baltic Region,5 pledging to improve skills development and access to data and to develop ethical guidelines, among others. Outside of Europe, the Arab AI Working Group aims to develop a joint framework for capacity building, train youth to compete in AI jobs and establish a common Arab Strategy, while the African Union has set up a working group to create a common AI strategy for Africa and to initiative projects in support of the Sustainable Development Goals (SDGs) (OECD, 2020[1]).

These commitments help to generate a collective vision for AI based on the unique contexts, cultures, norms and values of a region. They can also help each country bring its comparative strengths to the table, while filling in gaps elsewhere in the region. Additional benefits, depending on the structures of the agreements, can assist in achieving economies of scale for procuring digital solutions, contributing to efforts to share talent and knowledge. They can also pave the way for collaborative international projects and approaches. Such instruments can focus on both the impacts and considerations of AI for broad societal and economic goals, as well as objectives focused on innovation and transformation of the public sector itself.

At present, national governments across the LAC region do not have in place any shared strategies, commitments or other agreements to foster a joint vision and common approach to AI. Such a regional instrument could better align activities in the area, and would also reflect the OECD AI Principles, which highlight the critical nature of international co-operation as a key element for the successful development of AI (OECD, 2019[2]). While governments in the region have not established a collective vision for AI, it has demonstrated the ability to co-ordinate regionally on digital government issues through the Network of e-Government of Latin America and the Caribbean (GEALC Network), as discussed in Box 2.1. In addition, all governments within the scope of this review are member states of the Inter-American Development Bank (IDB), which has adopted the OECD AI Principles as part of its fAIr LAC initiative.6^,7

The OECD and the IDB have partnered to produce a data science toolkit8 for the responsible use of AI in public policy both within the LAC region and beyond. The report uses the AI system life cycle as a guiding framework to provide technical guidance for public policy teams that wish to use AI technologies to improve their decision-making processes and outcomes. For each phase of the AI system life cycle – planning and design, data collection and processing, model building and validation, and deployment and monitoring – the toolkit identifies common challenges related to the use of AI in public policy contexts and outlines practical mechanisms to detect and mitigate them (Sanchez Avalos, Gonzalez and Ortiz, 2021[3]).

This efforts demonstrate the region’s ability to collaborate on cross-border issues specific to AI, in spite of the fact that governments in the region have yet to agree on a regional approach to AI.

LAC governments have also demonstrated capacities and interest in regional AI collaboration on an ad-hoc basis. For instance, the IA-CKATÓN is a regional hackathon created to explore innovative ideas and novel ways of using AI to improve public services. Originally organised by Uruguay’s digital government agency, the Agencia de Gobierno Electrónico y Sociedad de la Información y del Conocimiento (AGESIC), the initiative expanded to involve Chile, the Dominican Republic, Panama, Paraguay and Peru. Each participating country carried out its own IA-CKATÓN and selected a national winner who participated in a regional grand finale at the annual meeting of the GEALC Network. Regional AI summits and conferences are another example of LAC regional collaboration. Two events are particularly noteworthy for attracting a diverse pool of actors from different disciplines and countries: the “Regional Forum on Artificial Intelligence in Latin America and the Caribbean”, organised by UNESCO and Brazilian partners in December 2019,9 and the “AI Latin America SumMIT”, organised by MIT Latin American researchers in January 2020. The latter recently published an e-book that documents the summit and provides details of planned future events.10 Such regional initiatives and networks represent remarkable advances in the promotion of AI in the public sector and understanding of its challenges and opportunities. Although still incipient, they represent a path towards greater regional collaboration at the strategic and policy levels.

LAC national AI strategies

While regional strategies can help guide collective action, the most comprehensive and granular strategies are found at the national level. At least 60 countries worldwide have adopted national AI strategies and policies to set strategic visions and approaches to AI (Berryhill et al., 2019[4]) (OECD, 2020[1]), with many others actively developing one. These strategies include AI-related priorities and goals and, in some cases, a roadmap for achieving them. They can help countries build a common foundation for success in their AI progress, as well as align the capacities, norms and structures of the relevant AI actors and ecosystems. The design of most national AI strategies underwent open public consultations and involved numerous stakeholders, including key industry consortia, academia, trade unions and civil society (OECD, 2020[1]). These efforts clearly demonstrate that many countries see AI as a national priority, and are willing to work openly with a broad array of stakeholders to build legitimacy and trust.

Among LAC countries, seven have developed, or are in the process of developing, a national AI strategy: Argentina, Brazil, Chile, Colombia, Mexico, Peru and Uruguay (see Figure 2.2).11 This trend signals a growing focus among LAC countries on ensuring they remain competitive with regional and global peers, and a determination to keep up with this rapidly evolving technology and its potential benefits and risks.

When LAC strategies are viewed collectively a number of key themes and objectives emerge. For instance, strategies often seek to catalyse economic development through R&D funding and incentives, transform the labour market and strengthen talent pipelines through upskilling programmes, and promote solid data governance and sharing, including through open government data practices. Notably, all strategies include provisions to help ensure that AI systems are designed and implemented in an ethical and trustworthy manner (e.g. the creation of ethical frameworks and governance bodies). A number of strategies also include a focus on international collaboration, notably the strategies of Argentina, Brazil, Chile and Peru. Some strategies include more specialised components, such as Chile’s inclusion of a gender-inclusiveness perspective in AI research and development. Most importantly for this report, all of the strategies include a specific focus on the use and implications of AI for public sector innovation and transformation, as discussed in the next subsection.

Most LAC countries, including those that do not have a current or forthcoming AI strategy, have published a broader national digital government strategy or related digital agenda or programme. These often include components that serve as foundational building blocks for AI (e.g. interoperability, infrastructure, analytics tools and processes, integration of services, etc.), although they do not generally incorporate AI as a main focus. Similarly, some countries have developed overarching data strategies (see the section on Foundational strategic data governance capacities), and while these include foundational elements for AI such as data sharing, they do not generally focus on AI-specific areas. However, there are indications that other LAC countries will shortly develop national AI approaches. Ecuador, for instance, has held discussions with academia, industry and civil society on joint efforts for AI development (Gómez Mont et al., 2020[5]). In Costa Rica, the government is collaborating with the IDB on a roadmap for a national AI strategy and an accompanying ethical framework (OECD, 2021[6]). In addition, the Dominican Republic and Panama have stated in interviews with the OECD that their governments have begun exploratory discussions about national approaches to AI, although plans for strategies have not yet been formalised.

Public sector components of national strategies

All of the seven countries with published or upcoming national AI strategies either have a separate strategy in place for AI in the public sector, or a dedicated focus embedded within a broader strategy (see Figure 2.2). This is critical as it allows AI to be integrated into policy making and service design processes.

Figure 2.2. LAC AI strategies and public sector transformation

Source: OECD analysis of AI strategies.

However, the strategies vary in the extent to which they demonstrate a focus on public sector transformation, and in some cases may no longer be in effect:

• Uruguay is the only LAC country with a dedicated strategy for achieving public sector transformation through AI.

• Argentina, Brazil and Colombia all have a dedicated focus on public sector transformation embedded within a broader completed strategy.12

• Peru highlights the public sector as a strategic focus area in its forthcoming AI strategy.

• Chile’s AI strategy and associated AI Action Plan highlight the importance of public sector AI training and adoption for enhancing public efficiency and service delivery, as well as streamlining public procurement processes. However, the general focus of the strategy is industry, with less attention overall to the strategic use of AI in the public sector than in other strategies in the region.

• Mexico’s strategy is not publicly available and thus could not be reviewed. However, subsequent reporting by those involved in its creation indicates that the strategy includes a significant focus on AI in the public sector. Mexico was the first country in the region to develop an AI strategy, but it is unclear whether it still constitutes active policy (see Box 2.2).

As with broader national strategies objectives, a number of key topics and themes and objectives emerge across the public sector-focused components that the OECD was able to review. Seven topic areas stand out as shown in Table 2.1.

Many of these topics and themes surface in the strategies of Uruguay and Argentina provided in Box 2.3.

Action plans and enablers for success

A comparative view of the strategies surfaces differences among LAC countries regarding the existence of action plans and enablers to help drive progress in implementation (Table 2.2). Although having these mechanisms in place does not guarantee successful implementation, they can improve overall performance, impact and accountability. In particular:

• All the reviewed strategies include objectives and specific actions, which is critical.

• Most also include measurable goals. For instance, Argentina, Chile (through an associated Action Plan) and Colombia present their goals in such a way that it is possible to measure progress over time. Uruguay’s strategy does not always include measurable goals, leaving some actions open to interpretation. Brazil’s strategy largely lacks measurable goals, with the exception of a goal to implement AI in at least 12 public services by 2022. Peru’s draft strategy, provided to the OECD for review in May 2021, provides many relevant objectives; however these are generally worded in a manner that does not allow for the measurement of progress or success.

• Chile and Colombia define responsible actors linked to each proposed action, which is important for ensuring someone or some organisation has ownership and accountability over implementation progress and success. Argentina defines responsible actors for each strategic axis but not for each action.

• Regarding the definition of time frames for the starting and completion of proposed actions, Colombia sets clear periods and Argentina defines time frames for certain actions. Chile includes time frames for initiating action for each action item, but does not include deadlines for completion. Brazil, Peru and Uruguay do not generally include specific time frames.

• Finally, Colombia’s strategy is the only one to include clear funding mechanisms (discussed further in the Funding section of Chapter 6) and a monitoring instrument (see Box 4.10 in Chapter 4).13

While only seven LAC countries have developed draft or final national strategies, and the OECD could review six in full, the themes, objectives, roadmaps and enablers discussed here can serve as a valuable reference for other countries as they assemble their own strategies. As mentioned above, a number of other LAC governments told the OECD in fact-finding interviews that although they do not currently have an AI strategy in place, they are laying the groundwork to build their own in the near future. With the recent launch of Brazil’s national AI strategy, the pace of development in the region appears to be increasing. The countries represented in this section have been regional vanguards from which others can learn, and who will also need to continually re-examine their progress and iterate and evolve on their own strategies to keep pace with technological advancements.

As additional resources in this area, the OECD.AI Policy Observatory provides access to a wealth of country-specific information on national AI strategy and policy initiatives.14

Ensuring strategy coherence and evolution

While the creation of national AI strategies is accelerating both in the region and globally, governments must be careful to ensure that such strategies form a core and integrated part of the country’s digital system. To be successful, these strategies must align with and mutually reinforce national digital government strategies (covered in depth in the forthcoming Going Digital: The State of Digital Government in Latin America), national data strategies (see “Foundational strategic data governance capacities” in Chapter 5), and ethical principles and values, and personal data protection policies and laws (see Chapter 4 on “Efforts to develop a responsible, trustworthy and human-centric approach”). Unless all these components work in tandem, public sector AI efforts will struggle to scale beyond small pilots and add public value. Like AI strategies, only a handful of LAC countries have developed comprehensive national data strategies. As such, governments in the region have an excellent opportunity to ensure coherence for these highly interrelated and interdependent strategies from the ground up. Those with strategies already in place can also achieve this convergence by ensuring their strategies are open to interaction and evolution.

Governments also need to ensure that their strategies are not one-off documents. AI is a rapidly evolving technology, and related strategies, policies and projects must also evolve in order to remain relevant and applicable. To help ensure that governments remain informed regarding the latest developments, countries inside and outside of the region have developed dedicated bodies to keep apprised of new information and to advise governments on how to respond (see examples in Box 2.4).

At the global level, governments are applying different models to ensure policy coherence and the effective implementation of national AI policies. These include:

• Assigning oversight of the development and implementation of a strategy to an existing ministry or agency.

• Creating a new governmental or co-ordination body for AI.

• Establishing AI expert advisory groups.

• Receiving input from oversight and advisory bodies for AI and data ethics bodies.

More details about these efforts can be found in the recent OECD report, State of implementation of the OECD AI Principles: Insights from National AI Policies (OECD, 2021[6]).