Colombia trains 404 university students in multi-agent AI to monitor low orbit and space

By Zendoric · September 29, 2026.

The Colombian Air Force and the Universidad de los Andes held a new edition of CODEFEST AD ASTRA, an innovation marathon that brought together university students and cadets from the "Marco Fidel Suárez" Military Aviation School around a specific challenge: to build a "Strategic Radar of Aerospace Trends," a multi-agent AI system — several specialized models collaborating with one another, each with a subtask — capable of scanning open sources to detect risks and opportunities in the military use of AI, the security of the space environment, the congestion of objects in low Earth orbit and the territorial dynamics of Latin America and the Caribbean, according to the Colombian Air Force's own statement.

The figure that gives a sense of the event's scale: 101 teams and 404 participants from 36 universities across 17 Colombian departments, of which 20 teams reached the in-person stage. The exercise, which began on May 20 with virtual training sessions and a qualifying challenge, closed on September 18 and 19 with 24 straight hours of work at the Universidad de los Andes. The air forces of Argentina, Brazil, the United States, Italy and Peru attended as observers; Chile's, by contrast, competed directly, the first in the region to do so in this format.

This piece should be read with the right filter: it is sponsored content, that is, an institutional press release distributed as such, with no third-party comment and no technical detail on which models or architecture the teams used, nor any information about the winning project. There is no operational product to assess here, but a training exercise. And that is precisely where it gets interesting.

In general, when we have analyzed AI's impact on employment sector by sector, the conclusion repeats itself: the winners are not those who use AI passively, but those who learn to orchestrate it to solve a real problem. This hackathon is exactly that: university students learning to coordinate AI agents for a use case that until recently was the exclusive preserve of intelligence analysts with access to classified tools. It is a route for building talent — and for future public-sector hiring — that deserves more attention than it usually gets amid the noise of the big frontier models.

There is also a governance angle we have touched on before when discussing AI as a strategic weapon and the dual use of these technologies: an OSINT (open-source intelligence) system with multi-agent AI applied to military capabilities and space security is, by definition, dual-use technology. That Colombia is training it in an open university hackathon, with international observers and growing regional participation, is a relatively healthy way to approach it: transparent, academic, collaborative. But the same pattern — AI agents scanning open sources to detect geopolitical risks — is the one we have seen used in other contexts for surveillance and disinformation. The difference is made by who controls the system and within what limits, not by the technology itself.

The low Earth orbit detail is no footnote either. The congestion of satellites and space debris is a real and growing problem on a global scale, and the fact that a Latin American university exercise includes it as a variable of analysis suggests that the region is starting to take space surveillance seriously, a domain that until now was monopolized by the major powers. That Chile is entering to compete, and not merely to observe, is a modest but concrete indicator of that trend: AI is lowering the cost of entry for mid-sized players into capabilities that once required superpower budgets. It is the same logic of democratization we celebrate when talking about open models, applied here to more sensitive terrain.