It's not whether you used AI, it's how: Universidad de Chile rolls out a protocol for training public managers

By Zendoric · October 1, 2026.

The School of Government and Public Management at the Universidad de Chile's Faculty of Government presented on Wednesday, September 30, its "Policy on the Critical and Ethical Integration of Generative Artificial Intelligence for Training in Public Affairs," together with a Protocol that brings it down to the classroom. The document grows out of a FADOP project led by the school's director, Lorena Oyarzún Serrano, and was built through surveys, focus groups and conversations with faculty and students in Public Administration and Political Science, as the university details.

The Protocol distinguishes four levels of generative AI use —preparation, review, minor delegation and major delegation— and translates them into a traffic-light system that indicates whether a given use is permitted, requires safeguards or is not allowed. The criterion is not binary ("did you use AI, yes or no?"), but relational: it depends on how that use relates to the competency the activity seeks to develop or assess, according to the document. The policy also sets an overarching principle —responsibility is not delegated to the tool— and includes data protection and gender equality among its guiding principles, because it recognizes that these systems can reproduce biases.

The urgency of the move has a figure behind it: more than 80% of first-time entrants to the Universidad de Chile say they use generative AI tools, according to figures cited by the director of the Undergraduate Department, Andrea Greibe Kohn, during the launch. Put another way: the policy does not get ahead of practice, it arrives after it. The Vice-Rector for Information Technology, José Correa Haeussler, went further and acknowledged that universities still do not fully know "how much AI can replace or, on the contrary, enhance" the capabilities they set out to build, and placed the debate within the broader public discussion about how Chile should regulate these technologies.

At Zendoric we have spent months documenting how AI is reshaping employment sector by sector, and education is no exception: the teacher who learns to orchestrate the tool within their discipline wins; the one who merely transmits content a model already generates just as well or better loses. What sets this policy apart is that it applies that same logic to those who will, tomorrow, take part in how the Chilean state itself uses AI —in the design and evaluation of public policy, in data analysis for public management. The school that trains future public managers is setting rules for itself first, and that order is no small thing.

It is also governance built the other way around from the usual: not a national law handing criteria down to a classroom, but a faculty that assembles, with surveys and focus groups, a framework that can then —in line with the responsibility Correa Haeussler attributes to the public university in that discussion— feed into the country's regulatory debate. It is the kind of evidence from practice that we defend against the risk of legislating out of panic: gradual, revisable criteria instead of generic bans that no one can enforce in practice.

The university itself treats this as a starting point, not a conclusion: the policy remains subject to review as the technology evolves. Its real value will be measured by whether it holds up as a living instrument or is shelved after the launch event. For a generation of public administrators entering their degree using AI on a massive scale, the line between using the tool and letting yourself be replaced by it will increasingly be the one that separates solid professional judgment from hollow judgment.