Sao Paulo has deployed Latin America’s most extensive artificial intelligence surveillance system in an ambitious attempt to combat the city’s persistent crime rates. The Smart Sampa program, launched in 2024, utilizes 40,000 cameras scanning the streets of Brazil’s largest metropolis, comparing real-time images against judicial databases to identify wanted individuals.
The system has demonstrated significant operational success, capturing approximately 3,000 fugitives and intercepting nearly 4,000 criminal acts in progress. Municipal Security Secretary Orlando Morando enthusiastically endorsed the program, stating the captured fugitives could fill seven prisons and that he ‘can no longer imagine Sao Paulo without Smart Sampa.’ The technology’s capabilities were demonstrated when Morando’s own image was tracked across multiple locations within seconds.
However, the implementation has revealed substantial flaws. Official transparency reports indicate that over 8% of arrests made through the system during its first year resulted in erroneous detentions. At least 59 individuals were released after being misidentified, including an 80-year-old retiree mistaken for a rapist and psychiatric patients interrupted during therapy sessions by armed police.
Further concerns have emerged regarding the system’s application beyond serious crimes. Nearly half of captured fugitives were wanted for ‘other’ offenses, predominantly child support payments—civil matters largely unrelated to public security. Critics argue the system enables ‘civil control’ rather than addressing violent crime.
Additional questions about potential algorithmic bias remain unanswered due to significant data gaps. Racial identity information is missing for more than half of those detained, creating uncertainty about whether the system exhibits racial bias in a country with one of the world’s largest Black populations. Studies from other nations have indicated facial recognition technologies frequently demonstrate higher error rates when processing darker-skinned individuals.
Despite these issues, government officials maintain the system has contributed to a nearly 15% reduction in robberies while deflecting responsibility for outdated warrants and missing racial data to the judiciary system.









