- Governing artifical intelligence (AI) is an operational challenge, not just a regulatory one. Success lies in building the structures, talent, and instruments to do so, an agenda supported by the IDB.
- State capacity means regulating, executing, and anticipating simultaneously. Spain built an institutional ecosystem with these three functions operating in parallel.
- Strengthening AI governance does not always require a specific law. Each country can combine regulatory instruments and public policy tools depending on its starting point.
How do governments build the institutional capacity to govern a constantly evolving technology?
Across Latin American and the Caribbean (LAC), several countries are already advancing in artifical intelligence (AI) strategies and legislation. El Salvador and Peru have existing laws; Brazil, Chile, and Colombia, among others, are working on regulatory proposals; and the Inter-American Development Bank (IDB) has proposed an enabling regulatory framework to balance innovation with rights.
But regardless of their level of regulatory progress, all countries face the challenge of developing institutional capacities to implement and adapt AI policies. Passing laws is not necessarily the first step. The real challenge is operational: equipping the State to supervise the development of AI systems, coordinate with technical actors, collaborate with the private sector and academia, and update its regulatory frameworks quickly.
Moving from paper to practice is hard. It requires iterating and course-correcting along the way. So rather than design theoretical models, the region should look at those who have already started. Spain's process is exactly this: a real-world case of how to turn legislative intent into an operational institutional ecosystem.
AI governance in Spain began as a state policy, part of a wider push to modernize the economy. In 2020, the Secretariat of State for Digitalization and Artificial Intelligence (SEDIA) was created within the Vice-Presidency for Economic Affairs, and the National Artificial Intelligence Strategy (ENIA) was published, backed by 600 million euros from the post-pandemic Recovery, Transformation, and Resilience Plan.
The structure kept changing after that. In 2023, AI governance became an independent Ministry for Digital Transformation and Civil Service to deliver the digital agenda. Later restructurings split AI from digital services and connectivity from the digital economy, and created separate directorates general for AI and for data governance.
For the region, the organizational chart matters less than what Spain built around it. First, a network of public entities, agencies, and investment companies gave the State the means to act, not just to regulate. These entities finance projects, run calls for proposals, channel investment, and provide technical services.Second, it created a deliberative and anticipatory layer comprising advisory councils and foresight, alongside multisectoral forums.
The result is an institutional ecosystem that integrates regulatory, executive, and anticipatory functions. A State that enables, rather than only regulates, may be the easiest part of the Spanish model for other countries to replicate.
In 2023, Spain became the first European Union Member State to create a supervisory body dedicated to AI: the Spanish Agency for the Supervision of Artificial Intelligence (AESIA). The agency became operational in 2024 and runs the first AI regulatory sandbox in Europe: a testing ground where companies run real AI systems under the regulator's eye to see how the rules would work, before they become law. It set up the agency before the European AI Act was even finished, on purpose: it wanted the institution ready before the law required one.
Spain's first national AI strategy, from 2020, was written before generative AI existed. So the updated 2024 AI Strategy had to add tools no one had in mind four years earlier, such as the language model ALIA and a 1.5 billion euro investment package. In a fast-moving field, the ability to update the vision matters as much as the vision itself.
No country has this fully worked out, and Spain is no exception. But its experience offers the region one example worth weighing: institutions can take shape alongside the rules rather than after them. The technology moves fast, and a law on its own rarely keeps pace, so what matters is having bodies that can enforce the rules and anticipate what comes next.
Four operational priorities LAC governments can start on:
- Start early, even if the structure is imperfect. Create or designate a lead AI authority with a clear mandate, budget, and coordination power. Institutional building can begin with existing frameworks (where they exist) or advance in parallel with the development of missing ones.
- Build the capacity to act, not just to write rules. This means public entities with a budget and a mandate, not only regulatory bodies. Without this layer, laws and strategies are rarely implemented.
- Attract talent with flexible hiring and competitive career paths. The shortage of skilled people is one of the most widely recognized obstacles to AI adoption in government, and without ways to compete for them, the institutions cannot hold.
- Leave room to revise as you learn. Periodic reviews, flexible mandates, and testing environments such as sandboxes let the State adapt, work with the private sector and civil society, and find the kind of regulation the country.
For Latin America and the Caribbean, where institutional frameworks vary widely, the same idea applies: build flexibility into institutional design from the start. The IDB works with countries in the region to strengthen the State for the digital era. Governing AI is not a single legislative act; it is a long-term institutional process. For more, see our work on this agenda.