Smart Sampa facial recognition system in São Paulo causes mistaken arrests
São Paulo's Smart Sampa facial recognition system, deployed in 2024 across 40,000 cameras to identify fugitives and criminals, reportedly caused more than 8 percent of arrests in its first year to be erroneous. At least 59 people were freed after being mistaken for others, including an 80-year-old retiree confused with a rapist and psychiatric patients arrested during therapy sessions.
What happened
Smart Sampa, Latin America's largest facial recognition system, was introduced in Sao Paulo in 2024 to scan images from 40,000 cameras and compare them to judicial databases to identify fugitives and people committing crimes.
Official transparency reports analysed by AFP show that more than 8 percent of people identified as fugitives and arrested in Smart Sampa's first year had to be released due to errors. At least 59 detainees were freed because the system mistook them for other people.
In December, an 80-year-old retiree spent hours under arrest because Smart Sampa confused him with a rapist. A month earlier, armed police handcuffed a psychiatric patient during a therapy session at a mental health centre after the system misidentified him; he was released after hours at the police station.
Reported impact
The mistaken arrests resulted in the wrongful detention and questioning of individuals, including elderly people and psychiatric patients. The Sao Paulo government stated that no one remained imprisoned by mistake, as people were released after being identified as mismatches.
Index analysis
The system relies on cameras in public buildings, including health centres, and private buildings that agree to participate. At least 141 people were arrested due to outdated warrants, though the Sao Paulo government argues those mistakes are the judiciary's responsibility.
Despite these errors, municipal security secretary Orlando Morando stated the program has captured 3,000 fugitives and nearly 4,000 people caught in the act of committing a crime, costing approximately two million dollars per month to run.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.