Imagine a military command center on the border. On a screen, an artificial intelligence program analyzes radar, camera, and satellite data in real time, recommending a strike within seconds. Officers decide whether to accept the recommendation. Yet the real danger is not on the screen, but thousands of miles away: the program runs on a foreign company’s servers, and no local judge or authority can review how it makes its decisions.
Delegating tactical intelligence for public defense to external infrastructure and software means surrendering the core of national sovereignty. While parliamentary debate on technology in Colombia focuses on sector-specific bills designed to regulate the use of artificial intelligence by security forces and restrict delegated lethal decisions, it overlooks a fundamental lesson from its regional neighbor: meaningful human control over military technology is impossible without industrial and technological sovereignty.
Industrial sovereignty and technological development
In the defense sector, framework laws prohibiting lethal autonomy become little more than dead letters if operational infrastructure is purchased from third parties as a closed package. The ability to audit, modify, and oversee technology requires a domestic industrial base. In this respect, the regional contrast is revealing.
Brazil has consolidated a defense industrial ecosystem led by multinational company Embraer, which reports annual revenues of more than $8 billion and manufactures military transport and attack aircraft such as the C-390 Millennium and the A-29 Super Tucano. The sector is complemented by companies such as Avibras, which specializes in advanced missile and artillery technology. This autonomy in hardware allows Brazil to integrate its own sensors and analytical systems into tactical platforms.
Colombia, for its part, has made significant progress in developing its own defense hardware. Through the Colombian Aeronautical Industry Corporation (CIAC), the country designs and assembles its own training aircraft and surveillance drones. In the maritime sector, the state-owned shipyard Cotecmar manufactures patrol vessels and river boats for the Navy, and reported exports of ships and floating structures worth $69.6 million in 2025. However, the remaining challenge goes beyond simply assembling steel hulls and aircraft fuselages: it requires learning how to develop, train, and inspect the artificial intelligence software operating inside them.
This is where Brazil offers a methodological lesson: the point is not simply to have better aircraft, but to ensure that, by controlling the platform, the state can require the competent authorities to audit the software operating it. Colombia is already building hulls and fuselages. The next step is to apply the same industrial logic to code.
Data infrastructure in Latin America
In the twenty-first century, sovereignty over public security rests on data infrastructure. An artificial intelligence model depends critically on the information used to train it and on the data center—the physical infrastructure where it processes operations. This is where the digital divide between Brazil and the rest of the region becomes crucial.
Brazil accounts for approximately 49% of Latin America’s data center capacity, according to Moody’s. Much of this capacity is commercial and foreign-owned, but on this foundation, the Brazilian state has promoted public cloud and supercomputing initiatives aimed at developing sovereign models. The difference between Brazil and Colombia is not merely one of scale: Brazil has a state policy designed to avoid dependence on external infrastructure.
Colombia, by contrast, lacks sovereign processing and supercomputing infrastructure. Delegating technological control compromises the state’s autonomy. The critical infrastructure used for the country’s analysis and decision-making remains structurally dependent on remote commercial servers located in the United States or Europe. If defense algorithms are trained on external data and run on foreign cloud systems that cannot be audited by local judges or military commanders, Colombia will have effectively privatized its strategic security.
Contrasting legislative approaches
At the legislative level, the two countries are taking different paths to address the threat posed by algorithmic lethality. In Brazil, Bill 2338 of 2023, approved by the Senate in December 2024 and still pending before the Chamber of Deputies, established a regional precedent by requiring any artificial intelligence system that poses risks to life or physical integrity to be subject to human supervision and a final human decision. In doing so, it establishes a standard of safeguards for public auditability in the face of algorithmic autonomy. However, this model risks remaining a set of abstract principles unless the military operational framework establishes clear rules. The legislative deadlock demonstrates that even Brazil has yet to resolve this debate.
Colombia, meanwhile, has begun processing initiatives targeting its security forces, such as House of Representatives Bill 364 of 2026. But the country cannot reduce the debate to a simple formal prohibition that ignores the urgent need to develop its own technological capabilities. Focusing regulation on prohibiting autonomous firing creates the illusion of ethical control, when in reality, abandoning this technological race leaves the country at a military disadvantage and keeps its defense intelligence tied to foreign servers.
Toward regional technological autonomy
To build effective regulation that protects life and human rights, Colombia must combine an operational approach with the principle of digital sovereignty that Brazil is promoting. A robust regional regulatory framework requires moving beyond passive imitation of foreign rules and incorporating structural safeguards into domestic law.
The rule is clear: an algorithm must never decide matters of life and death. Every analytical system used by the armed forces must remain subject to the supervision of an operator or military commander who is accountable before the ordinary courts.
Furthermore, it is essential to require that national intelligence and security data, as well as the training history of military analytical models, remain on sovereign servers within the country’s territory. This must be complemented by public records and civilian audit mechanisms to verify that the software being used complies with international humanitarian law.
A state that does not control the software, models, or servers used to analyze threats to its security is surrendering its sovereignty.
Autor
Lawyer and professor of Human Rights and Technological Power in the International Order at the Universidad de los Andes, Colombia. Specialist in international human rights, with a focus on artificial intelligence (AI) governance, technology policy, and responsible AI.
