Brazil

What the record shows

Brazil runs its cities through contracts, and its record in this atlas is mostly a record of what those contracts do and do not oblige. The technology is rarely the variable. Two cities put nearly identical bikes on nearly identical docks and got opposite arrangements: São Paulo grants a revocable permit under which the operator is paid in advertising rights, so the sanction for an undelivered fleet is fewer logos rather than breach of contract, while Rio sells the street for cash — R$17.5M up front in 2017, then a competitively tendered concession in 2022. Payment bought Rio an expansion target and nothing else: the city still owns neither the bikes, nor the docks, nor the data.

The same pattern runs through the lighting retrofits, which are the country's most repeated smart-city purchase. IluminaSP converted more than half a million points to remotely managed LEDs and finished two years ahead of its contract. Belo Horizonte's twenty-year PPP, reportedly the first streetlight partnership operationalised by a Brazilian state capital, reached full LED coverage in three years — and its own concessionaire's account puts the remote-control layer on about 32,000 of the network's 186,000 points, which is the number to hold onto before crediting a whole city as smart. Curitiba's shows what happens when a concession slips: the LED deadline was missed, the state audit court found service shortfalls, the city fined the operator, and in June 2026 the concessionaire agreed to sell the contract on.

Campo Grande is where the contractual logic turns into the sharpest story in this record, and it turns on a monitoring system rather than a sensor failing. Every bus in the city has reported its position since 2016, and the regulator reads that feed backwards: an unrecorded trip is an omitted trip and draws a penalty notice. A special municipal commission counted 21,910 of them between 2021 and 2025, and in June 2026 the city placed the concession under intervention and removed the consortium's management. The uncomfortable detail is in the same report — the GPS units, cameras and turnstiles that generate the compliance record were frequently defective, and they belong to the party being measured. A missing trip and a broken tracker look identical in the data.

Brasília does the same thing deliberately and writes the tolerance down. SIMLUR does not merely track the cleaning contractors' vehicles, it pays them from the track, overlaying the day's satellite positions on the planned rounds and counting the kilometres actually covered — with an explicit five per cent allowance written into the instruction for signal interference. It is an unusually candid piece of drafting, and like almost everything else here it has no published evaluation of whether measuring the contractor this way improved the service or only the paperwork.

That absence is the country's most consistent feature, and it is most visible in the building Brazil is famous for exporting. Rio's Centro de Operações opened at the end of 2010, months after April rains killed nearly seventy people, as a partnership with IBM, and it became the template: sixty-plus agencies and thousands of cameras in one room. Belo Horizonte followed in 2014, Manaus in 2020, Campo Grande in 2022, Recife in 2023, Curitiba in 2024. Camera counts and investment figures are easy to find for all of them. Independent measurement of what any of them changed is not, and Campo Grande's centre is honest about the gap: its most consistent measurable output is a maintenance queue.

Where the record is unambiguous is on participation, and it is not a happy reading. Porto Alegre invented the participatory budget in 1989 and hundreds of cities copied it; at home the assemblies stopped taking new demands after 2016, and the share of municipal spending it touches fell from all of the investment budget in the mid-nineties to a fraction of a per cent. Belo Horizonte's OP Digital was among the first digital participatory budgets anywhere and lost three-quarters of its voters between its first and third cycles before quietly not being renewed. The lesson the Porto Alegre entry draws is the country's, not the city's: the world's most-copied democratic innovation was never protected by law, so when the party that built it lost power the successors kept the name and cut the money.

One last thread is worth following because it shows a limit that is legal rather than technical. Campo Grande bought cameras with plate reading and face matching in the same order. The plate half issues automatic alerts. The face half has never been switched on, because the municipal guard holds no image database of its own and drawing on police records under Brazil's data-protection act requires a formal agreement nobody has signed — while the state government runs facial recognition at a trade fair in the same city. Belo Horizonte's answer to the same problem is to send its feeds to a federal platform, which moves the accountability question to an agency the city does not answer to; in Rio the biometrics belong to the state military police, which reports arrests and no error rate at all. Three cities, three different places where the capability stops — and in only one of them was the reason the equipment.

Sources beyond the atlas

Facts above that do not come from a linked entry rest on these:

• Lei Geral de Proteção de Dados Pessoais (Law 13,709/2018) — the data-protection statute behind the agreement Campo Grande's municipal guard has not obtained
• Lei 8.987/1995, the concessions act — the legal basis for the intervention in Campo Grande's bus concession