On a crisp winter’s morning in June in Mata de São João, fourth graders hopped off the bus onto the dusty track in front of João Pereira Vasconcelos school. It had been a long two-year break from the classroom due to Covid-19, but as the students filed in through the school’s run-down entrance, they received an unexpected welcome. Their school had become the latest test pilot for facial recognition cameras to control school attendance.
In April, João Gualberto, the district mayor, held an in-person auction letting Brazilian technology companies bid for a contract to supply facial recognition technology for the public school system in Mata de São João.
The 900,000 reais (about $162,000) tender was won by PontoiD, and in July, two public schools — João Pereira Vasconcelos and Celia Goulart de Freitas — began discreetly rolling out the facial recognition system, without informing parents or students in advance.
Students were registered on the system, which built a high-resolution map of their faces. Now, when they enter the school, the cameras scan and register them. A text message is then automatically sent to parents, directors, and school staff to alert them when students arrive and leave school.
Alex Carvalho, the district’s secretary of education, said the technology will improve attendance and safety, help to digitize report cards, and inform kitchen staff exactly how many people they need to prepare meals for. As an added bonus, the technology has also recently doubled as temperature tracking, acting as a Covid-19 preventive measure.
After testing in two schools — the results haven’t been disclosed to the public — authorities have now started rolling out the technology across the district’s dozen remaining public schools.
While much of the world is debating the dangers of biometric surveillance and digital invasions of privacy, Brazilian officials of all political stripes are welcoming facial recognition technology with open arms. Its rollout in public schools reflects the complex considerations citizens and politicians must make in a country with high crime rates. For many in Brazil, submitting their biometric facial data is a price they feel they must pay for public security.
But despite politicians’ strong interest in modernizing schools and enhancing security, private activists warn that parents are often unaware of the risks involved. For skeptics, the unchecked use of high-tech tools without much legal protection as well as a lack of transparency and accountability surrounding the deployment of these technologies are major concerns.
In Mata de São João, a small, low-income tourist hub 60 kilometers north of Salvador, the capital of Brazil’s northeastern state of Bahia, people have greater reason to be worried. The town was one of the first slaving ports in the Americas, and, today, more than 80% of its roughly 50,000 people are Afro-descendant, a historically low-income population that has faced systematic discrimination in Brazilian society.
For cybersecurity activists, deploying biometric tools on a population that often has little say in the matter could reinforce preexisting racist practices from authorities.
Luciano Melo, a 37-year-old bus driver and parent at Valdete Seixas public school, said that some parents in Mata de São João were more concerned than others. He told Rest of World that he had been caught off guard when a facial recognition system began registering his 11-year old son last week without his consent. “No parent here has been informed,” he said. “It shouldn’t be this way.”
Brazil’s president Jair Bolsonaro was elected in large part on a pledge to crack down on crime, triggering a surge in mass surveillance ostensibly for public security in cities across Brazil. Soon after coming to office in 2019, the former army captain signed a decree to digitize crime stats, compelling all federal bodies to share the vast troves of data they gather on Brazilian citizens and consolidate it in a single centralized database. The surveillance system in Bahia also uses the National Bank of Arrest Warrants.
Across the country, cities and states have begun to mull a patchwork of bills that could make crime-fighting technology, like facial recognition, obligatory in public spaces.
Schools are just the latest front, acting as what experts consider an integrated extension of police mass surveillance as a key crime-fighting tool — in this case, aimed to help prevent violence against children and other youth crimes in the neighborhood.
PontoiD, self-described “pioneers” of facial recognition software in Brazil’s public schools, currently operates in 19 of Brazil’s 26 states and, in line with a nationwide push to adopt the surveillance, has recently ventured into remote peripheries of the country, including Mata de São João.
Bahia isn’t Bolsonaro territory. It is run by Rui Costa, a progressive, left-wing governor from the Workers’ Party (PT), but his state government has enthusiastically embraced surveillance technology.
Costa has a strong relationship with China — a microcosm of the relationships constructed by the PT over the past decade when it was in federal office. Over the years, Bahia has received a series of Chinese investments across sectors, including infrastructure, solar energy, and facial recognition technology — often at rates 10 times cheaper than their competitors’, sometimes even for free.
In 2018, tests were then carried out in carnival blocks in Salvador using facial recognition cameras bought in part from telecoms giant Huawei for 18 million reais (about $3,236,770) before Costa shared the results at a conference attended by representatives from Huawei on a trip to the Chinese city of Shenzhen.
“No parent here has been informed. It shouldn’t be this way.”
In July this year, Costa signed a $131 million partnership with Brazilian telecom company Oi solutions, which were contracted to operate and monitor Huawei’s cameras after the initial testing in Salvador. The plan is now to expand the facial recognition cameras to 77 cities across the state.
Bruno Sousa, a Bahia-born researcher at O Panoptico, an independent facial recognition surveillance body at the Candido Mendes University Center for Studies on Public Security and Citizenship (CESeC) in Rio de Janeiro, said the trend is worrying, given the preexisting racial and gender biases in both the current available technology and the police force.
The Brazilian police rarely disclose publicly the circumstances of arrests, so O Panoptico has taken to directly requesting interviews with public security departments and filing freedom of information requests to analyze data from arrests across Brazil.
“We have already seen higher error rates [when it comes to surveilling] Black faces and an absurdly high rate of false positives,” said Sousa.
Data compiled by researchers at the Security Observatory Network found that 90.5% of those arrested in five Brazilian states using facial recognition were Black. Worrying news for Bahia, which is home to the largest Black and pardo (someone of mixed ethnic ancestry) population outside Africa.
While current facial recognition technology easily differentiates between the faces of white men, research has shown a disproportionately high error rate on Black faces — particularly Black women’s faces. That’s because a large part of the database of images used to train the algorithm is dominated by white, male faces.
PontoiD Technology’s commercial director, Sandro Bio told Rest of World: “We have shown a very high accuracy on all Brazilian ethnic skin types.”
Other supporters of surveillance technology argue that the consequences of false matches are minimal: A simple ID scan will prove that the suspect’s identity does not match that of any criminals on the police database, and they will be released.
But that underplays the risks of arrest in a country where police impunity is rife and where Black people are often disproportionately discriminated against, according to Pedro Diogo, a lawyer who researches surveillance technology and racial terror at the Federal University of Bahia and consulted on the article.
Diogo cited a recent case of a 25-year-old special needs man who was falsely identified by AI cameras and arrested by police at a Bahia bakery. The police put a gun to his head but released him when his mother confirmed his identity.
