Duplicate Image Crisis Hits São Paulo's Digital Public Records: The Key Decisions Ahead

São Paulo's municipal technology directorate is under pressure to act after an internal audit identified thousands of duplicate image files embedded across official urban planning and property registration systems — redundant photographs, permit scans and satellite snapshots that are inflating storage costs, slowing query times and, in at least some cases, generating contradictory administrative records for the same address. The problem is not new, but the scale of it, now formally documented, has forced the question of who fixes it, how, and by when.

The timing matters. Mayor Ricardo Nunes has staked part of his second-term agenda on digitising city services, particularly the licensing and zoning workflows that run through the Secretaria Municipal de Urbanismo e Licenciamento on Rua São Bento, in the historic centre. If duplicated images are allowed to persist in those workflows, decisions about building permits in dense neighbourhoods like Brás or Mooca could rest on conflicting photographic evidence — a serious liability in a city where urban litigation is common and expensive.

What the Technical Fork Looks Like

Engineers working inside the city's GeoSampa platform — the municipal geographic information system maintained by PRODAM, the city's public IT company — face a fundamental choice: run a perceptual-hash deduplication pass across the entire image archive, or build a slower, more expensive layer of human-verified curation starting with the highest-stakes records first. The first option is faster and cheaper but risks flagging legitimately distinct images — two photographs of the same building taken years apart, for example — as duplicates and deleting one. The second option could take until well into 2027 to complete at current staffing levels, according to the nature of similar projects undertaken by other large municipal governments.

PRODAM has handled large-scale data migration before. In 2023 the company managed a migration of more than 40 terabytes of property cadastre data as part of an upgrade to the city's IPTU assessment system. That project took roughly eight months from planning to sign-off. A deduplication exercise of equivalent complexity would likely follow a similar timeline, which puts any clean solution well past the end of 2026 — assuming a decision is made soon.

Private-sector players are circling the contract. At least three São Paulo-based tech firms with addresses in the Vila Olímpia and Faria Lima corridor — the city's de facto enterprise software district — have existing framework agreements with PRODAM that could, in principle, be extended to cover deduplication tooling without a full public tender. Whether the city chooses that route or opens a new competitive bid will be one of the first concrete decisions to watch.

The Policy and Legal Pressure Points

Brazil's Lei Geral de Proteção de Dados, in force since September 2020, adds a legal dimension that is easy to underestimate. Where duplicate image files contain identifiable personal data — photographs attached to housing assistance applications processed through the Secretaria Municipal de Habitação, for instance — retaining unnecessary copies may constitute a compliance breach. The Autoridade Nacional de Proteção de Dados has the authority to investigate municipal entities, and a formal inquiry would turn a technical housekeeping problem into a public accountability story fast.

Community groups in districts like Heliopolis, where digitised housing applications have been part of the regularisation process for thousands of families, have a direct stake in whether their records are correctly deduplicated or accidentally altered. An error that merges two residents' files because their address photographs were flagged as identical could disrupt regularisation processes that have been running, in some cases, for more than a decade.

The next 90 days are when the critical choices get made. The city must decide on a technical approach, determine whether to use existing vendor contracts or run a new tender, assign a named lead within PRODAM or the urbanismo secretariat, and publish a timeline. None of those steps requires new legislation. They require a decision. Until one lands, the duplicate images — and the contradictory records they generate — stay exactly where they are.