Rafael Santoro
The next president will inherit more than a security crisis. He will inherit an expanding architecture of observation.
São Paulo, Brazil
Brazil is voting for a president, but something less visible is also being decided. Crime, public order and institutional authority have returned to the center of political debate, while cameras, facial recognition, artificial intelligence and interconnected databases are becoming ordinary instruments of government. None of these technologies appears on the ballot, yet their future will be shaped by decisions made long after the campaign banners disappear.
Surveillance rarely arrives as an ideology because it normally presents itself as a practical answer to an immediate problem. A camera can identify a suspect, an algorithm can accelerate an investigation and an integrated database can connect institutions that once operated separately. In a country confronting organized crime and persistent urban violence, refusing such capabilities can sound detached from everyday insecurity. The legal difficulty begins when an exceptional tool becomes ordinary infrastructure.
Brazil already offers evidence of that transition. Facial-recognition systems have entered public-security operations in several states, allowing authorities to compare images captured in public spaces with databases used by law-enforcement agencies. Supporters point to arrests and investigative efficiency, and those outcomes cannot simply be dismissed. They also do not answer what kind of institutional architecture is being constructed around the technology.
That architecture determines how biometric information enters a database, how long it remains there and what degree of human verification precedes police intervention. An incorrect match is not merely a technical error when it can place an individual inside the machinery of criminal enforcement. Accuracy therefore matters, but legality cannot be reduced to a percentage displayed on a technical report.
Brazilian lawmakers have begun confronting some of these questions through proposals governing artificial intelligence in public security. Human supervision, restrictions on automated decisions and judicial authorization can establish meaningful limits, although regulation often develops after systems are already operating. Administrative normalization can occur quietly, leaving law to define boundaries around practices that institutions have learned to regard as routine.
The electoral dimension makes this particularly important because surveillance infrastructure is unusually resistant to political alternation. A network installed under one administration can be expanded by another, and a database created for a narrow purpose can later become useful for something never contemplated when it was established. Governments change faster than technological systems, especially when those systems begin demonstrating operational value.
Reducing the matter to Lula or Bolsonaro would therefore obscure the more durable institutional question. Governments with very different political orientations can find the same appeal in technologies that promise greater visibility over difficult urban environments. Surveillance does not need an explicitly authoritarian project in order to expand. Administrative convenience, public anxiety and bureaucratic habit may be sufficient.
Brazil’s megacities complicate the equation further because surveillance is not experienced uniformly across urban space. Avenida Paulista, a gated condominium and a heavily policed peripheral neighborhood exist under different relationships with state authority. An algorithm may be technically identical in all three locations while operating within social environments marked by very different levels of police exposure and access to legal protection.
The private sector introduces another layer that electoral debates rarely examine closely. Security platforms, biometric systems and analytical software are frequently developed or maintained by companies whose proprietary technologies may be difficult for public agencies, courts or citizens to inspect independently. When an automated process contributes to an erroneous detention, responsibility can become distributed among officers, administrators, contractors and software providers.
That fragmentation creates a particularly difficult problem for law. Rights become meaningful when there is an identifiable institution capable of answering for their violation, while complex technological systems can disperse responsibility across several actors without eliminating the harm itself. Operational sophistication may therefore increase at the same moment that legal accountability becomes harder to locate.
Brazil enters the runoff carrying a security debate much larger than competing promises about crime. The next administration will inherit an expanding capacity to identify, correlate and monitor individuals across physical and digital environments, together with legitimate public pressure to use those capabilities effectively. The unresolved question concerns the limits that remain necessary when tools designed for specific threats gradually become part of ordinary government.
There may never be a single moment at which Brazil decides to construct a surveillance state. Such systems are more often assembled incrementally, through technologies that appear reasonable when considered individually and become politically significant only after they are connected. The election will decide who governs Brazil next, but the architecture being built around security may remain long after that answer changes.
Rafael Santoro | Senior Columnist, Phoenix24 | Cybercrime, Surveillance & Urban Techno-Law