A São Paulo resident has been wrongfully detained four times in seven months after being misidentified by the city's Smart Sampa facial recognition surveillance system, raising urgent questions about the accuracy and oversight of biometric policing technology in Brazil's largest metropolis.
The case, documented by G1, involves a man from the city's South Zone who was repeatedly flagged by automated cameras as matching individuals wanted for criminal offenses. Each time, he was detained by police, questioned, and eventually released after officers confirmed his identity—only to have the same scenario repeat weeks later.
In Brazil, as across Latin America's giant, continental scale creates both opportunity and governance challenges. São Paulo, a megacity of over 12 million people, has rapidly expanded its use of facial recognition technology as part of the Smart Sampa program, which deploys thousands of cameras across the metropolitan area to identify suspects in real-time.
"Every time I leave my house, I worry the police will stop me again," the victim told G1, requesting anonymity due to fear of further harassment. "I've lost work because of these detentions. My family is traumatized. And the system keeps making the same mistake."
The Smart Sampa system, launched in 2019, uses artificial intelligence to compare faces captured by street cameras against a database of individuals wanted by law enforcement. Municipal authorities have promoted the technology as a breakthrough tool for public safety, crediting it with hundreds of arrests.
However, civil liberties advocates have long warned that facial recognition technology suffers from significant accuracy problems, particularly when identifying people of color. Studies in the United States and Europe have documented error rates up to 10 times higher for Black and brown faces compared to white faces—a disparity that appears to be manifesting in 's system.
