How LATAM’s Leading Countries are Guiding AI Development

AI technology might be too powerful –and appealing– to be contained. That doesn’t mean that government regulators aren’t trying.

Even since the explosion of generative AI (GenAI) into the mainstream, regulators the world over have tried to set up guardrails around the development of AI models. Most of them aim to keep the technology away from nefarious players and avoid the normalization of practices that might result in the mishandling of personal data, IP theft and even human rights abuses.

The European Union stands today as perhaps the most advanced territory when it comes to AI regulation. In mid March, the European Congress approved a comprehensive set of “risk-based” regulations that force organizations developing powerful AI models to comply with certain transparency and copyright laws. Riskier models are expected to stick to tougher rules.

Regulation doesn’t always lead to stricter rules, though. In India, for example, government authorities recently pulled back what should have been the country’s most comprehensive and tough regulatory framework for AI. Following industry pressures, regulators tweaked the new rules to make them less burdensome and allow for AI technology to be developed more “freely”.

In the American nearshore, the situation isn’t as clear cut. Most of Latin American more relevant markets for the IT industry have yet to implement solid and clear regulations on the development of AI models. 

This article seeks to provide a quick rundown of the state of AI “guardrailing” in the nearshore; specifically in Mexico, Argentina, Brazil and Colombia. Some of these territories are still figuring out what would be the best way to handle the technology. In other words: what follows is a survey of the state of things as they are today. They might change in the near future.

Argentina

Argentina –despite its economic troubles– is among the most advanced Latin American countries when it comes to AI regulation.

In mid 2023, Argentina’s Secretariat of Public Innovation published a set of guidelines and regulations for “fair and ethical” adoption of AI. These guidelines –named “Recommendations for a Trustworthy AI”– aren’t mandatory, however. They seek to push developers of AI algorithms towards a human-centered approach to their task; that is, AI should be developed ethically and should always operate with human guidance.

“This guide seeks to incorporate ethical and transversal principles into the whole cycle of AI development, defining risks and responsibilities every step of the way,” stated Agustina Brizio, Argentina’s Undersecretary of Information Technologies, when the regulations were unveiled. “We’re prioritizing a focus that guarantees respect for human rights, democracy and a gender lens; this while promoting governance and close cooperation between the private sector and academia to produce an innovative ecosystem.”

The recommendations include: not forcing AI as a solution to all sorts of problems; keeping human supervisors around the technology; proper treatment of training data; AI models being as transparent and comprehensible as possible; safe and “auditable” implementation of such models; and the participation of diverse and multidisciplinary teams, aware of ethical challenges, in all AI projects. 

As of August 2023, there were three legislative bills in Argentina advocating for AI regulation. None of them have been approved. 

President Javier Milei took office in December 2023. He has yet to take an official stance on AI development. He is known, however, for his extremely hands-off approach to government, particularly in matters of economic policy and business development.. 

Brazil

Despite its reputation as Latin America’s premier destination for tech development, Brazil finds itself in a transitional period in regards to its regulation of AI algorithms.

In November 2021, Brazil’s Congress approved a bill (PL 21/2020)  that regulates AI development in the country. Analysts regard it, however, as more “business friendly” and “hands-off”, opting for a decentralized system which gives the government very little power to oversee AI projects. 

Another bill is making the rounds in Congress, however. Introduced in May 2023, the new bill (PL 2338/2023) has more of a European approach to AI regulation. That is: it is more restrictive, and aims to implement guardrails around AI projects according to their level of risk. 

The bill requires, among other things, the prompt reporting of security incidents related to the AI algorithm. It will also give the Executive Branch the power to decide which government agency will oversee development.

There has been enough legislative attention given to AI regulation in Brazil to lead one to believe that another bill will eventually emerge. At the moment, however, AI developers have ample room to operate in the country, with relatively little oversight.

Mexico

Surprisingly, Mexico lacks any AI-specific strategy at a government level, and there are no specific laws for AI development either. The federal government –under President Andres Manuel Lopez Obrador– has paid little attention to tech development, focusing almost exclusively on Internet connectivity. Lopez Obrador’s administration is coming to an end this year, however, so that might change pretty soon.

There are four laws in Mexico which regulate “AI-adjacent” issues at a federal level: the Federal Law on Broadcasting and Telecommunications; the Federal Law on IP Protection; the General Law on Personal Data Protection; and the Law for Economic Competition. There’s also a bill for a Federal Cybersecurity Law that has yet to be discussed in Congress.

In May 2023, the Mexican Senate created an “AI alliance” with businesses and other stakeholders to regulate AI. While not a law or regulation, strictly speaking, it points to some political will in the country to make things happen. 

And given Mexico’s rising prominence as an exporter of tech services to the US, things should be happening sooner than later.

Colombia

Colombia once was one of the world’s more advanced countries when it came to AI readiness. Back in 2021, the country ranked 23rd in Tortoise’s AI Global Index, which measures each country’s readiness through the availability of talent, R&D credentials, infrastructure, operating environment, etc. In the latest Tortoise Index, Colombia ranked 48th.

Much of the efforts implemented by former Colombian administrations have been halted in the current government of President Gustavo Petro. His administration has given relatively little attention to AI development, particularly in matters of R&D. 

However, Congress is actively pursuing regulation. Four bill proposals have been introduced in August and September of 2023 alone. 

Authorities have also published an ethical guideline and a regulatory sandbox on AI. While they suggest more than enforce, both documents point to a understanding that there’s a need for ethical regulation.