São Paulo Lost Thousands of Historical Images Through File Corruption

The problem has a name inside the Prefeitura de São Paulo's IT directorate: imagem fantasma, ghost image. It refers to the thousands of duplicate, mislinked, or corrupted image files that have accumulated across municipal web portals since at least 2018, producing the blank grey boxes and broken-link icons that anyone who has tried to navigate the city's official transparency portal, SP Transparência, knows intimately. This week, a working group inside the Secretaria Municipal de Inovação e Tecnologia confirmed it has begun a phased audit of the problem, a quiet admission that years of deferred maintenance have compounded into a structural crisis for public digital records.

The timing matters. Brazil's Lei de Acesso à Informação, the federal access-to-information law enacted in 2012, obligates all levels of government to maintain legible, accessible public records. When images attached to contracts, urban planning approvals, or public health bulletins simply disappear, replaced by duplicate placeholder files or dead URLs, that is not just an aesthetic failure. It is a compliance problem. Journalists, lawyers, and civil society organisations working along Rua Líbero Badaró in the city centre have documented case after case where missing visual records made it impossible to reconstruct the timeline of a licensing decision or a public works contract.

How the Backlog Grew

The roots of the current mess stretch back to a 2017 migration of several city platforms onto a unified content management system that was never fully completed. Subsequent administrations layered new portals on top of old ones without retiring the legacy databases. By 2022, the city's main urban planning portal, GeoSampa, was hosting at least three separate versions of certain neighbourhood map images, some updated, some outdated, some simply duplicates of each other with different file names. Engineers who worked on the system at the time described the situation in public procurement documents, which noted the existence of redundant file trees without specifying how many files were affected.

The practice of uploading duplicate images accelerated during the Covid-19 pandemic, when multiple city departments rushed to digitise physical records without a common naming protocol. The Arquivo Histórico de São Paulo, housed in the Edifício Conde de Prates near the Pátio do Colégio, flagged the problem internally in a 2021 technical report, a document later obtained by civil society group Transparência Brasil through an information request. That report estimated that roughly 30 percent of newly digitised files in one departmental batch had been uploaded more than once, with no deduplication step built into the workflow.

The financial cost is harder to pin down, but it is not trivial. Cloud storage for redundant files, staff time spent manually verifying records, and legal exposure from incomplete archives all factor in. The municipal IT budget for 2025 allocated R$ 47 million to digital infrastructure maintenance, a figure drawn from the city's published Lei Orçamentária Anual, yet auditors at the Tribunal de Contas do Município de São Paulo noted in a March 2026 report that recurring maintenance of legacy content had not been itemised as a discrete line.

What Comes Next

The current audit, expected to run through October 2026, is using open-source deduplication tools alongside a protocol developed with the Instituto de Pesquisas Tecnológicas, the state-linked research body based in Cidade Universitária. The goal is to identify canonical versions of disputed files, delete confirmed duplicates, and implement a redirect system so that external links, including those embedded in years of journalistic reporting, resolve to the correct image rather than a broken placeholder.

For ordinary paulistanos trying to track a zoning change in Pinheiros or verify a public health contract in Ermelino Matarazzo, the practical advice is blunt: if a document you need shows a broken image, file a formal pedido de acesso via the SP156 portal and request the physical or original digital file. That workaround is slow, the city has 20 business days to respond under federal law, but it remains the most reliable path to a complete record while the audit runs its course. The ghost images are not going away overnight.