AI cameras promise safety. The bill arrives in rights
Facial recognition in the streets, social media monitoring, "predictive" policing: AI surveillance is advancing in Latin America faster than the laws that should limit it. It is neither science fiction nor neutral: it fails more often on brown and Indigenous faces, it gets used against journalists and land defenders, and it almost never answers to anyone. Understanding how it works is the first step to demanding rules.
What is being installed, exactly
Three technologies are spreading through the region’s cities, often purchased without public debate. Facial recognition: cameras that compare every passing face against databases — sometimes of suspects, sometimes of all citizens. Social media monitoring: systems that track what is posted about certain topics and who posts it. “Predictive” policing: programs that say where to patrol based on past crime statistics.
Where the problem lies
First, the errors are not evenly distributed. Facial recognition systems are trained mostly on faces from other latitudes: they fail more on brown skin, on Indigenous features, on women. A “false positive” is not a statistic — it is a person detained for a crime they did not commit, and such cases are already documented.
Second, the crosshairs point crooked. In several countries of the region, these tools have been used not against crime but against journalists, opposition figures and land defenders. For a community protesting a megaproject, “security” can mean being watched.
Third, prediction inherits injustice. If police always patrolled poor neighborhoods more, the data says more crime happens there, so the system sends more patrols there. Prejudice enters dressed as mathematics.
Not every use is equal
Neither panic nor hype: AI also finds missing persons, detects child abuse imagery and helps in disasters. The difference between legitimate use and abuse is not set by the technology — it is set by rules: is there a law authorizing it? A court order? A public registry of which systems the government uses? A path to appeal if it affects you? Where those answers are “no,” the tool rules over the right.
What you can demand
At minimum: transparency (which systems your government bought, from whom, with what results), limits (banning mass facial recognition in public spaces, as several congresses are already debating), independent auditing (someone external measuring the errors and against whom they fall) and remedies (the power to challenge a decision made by a machine). Surveillance installs itself in silence; rights are defended out loud.