'My Whole History, Gone': SF Residents Speak Out as Duplicate Image Removal Sweeps Community Archives

Dozens of San Francisco residents say they have lost irreplaceable photographs from shared community digital archives over the past three months, after a wave of automated duplicate-detection tools flagged and removed images stored on platforms used by neighborhood associations, nonprofit organizations, and local history projects across the city.

The deletions hit hardest in the Mission District, the Tenderloin, and the Western Addition — neighborhoods where community organizations have spent years digitizing fragile records of displacement, protest, and cultural life. For many residents, the loss is not a minor technical inconvenience. It is the erasure of documentary evidence tied to ongoing housing and eviction fights.

The issue is urgent right now for a specific reason. San Francisco's Department of Technology has been pushing city-adjacent nonprofits to migrate data to consolidated cloud infrastructure since January 2026, a process that has exposed legacy image libraries to new deduplication algorithms they were never designed to survive. Community groups that accepted small technology grants from the city to fund the migration say they were not warned their archives would be subjected to automated content review during the transfer process.

Organizations on the Front Lines

Two organizations have become focal points for the frustration. The Mission Cultural Center for Latino Arts, on 24th Street in the Mission, runs a digital archive dating to the early 1990s that documented neighborhood murals, many of which have since been painted over or destroyed. Staff members say the archive lost a significant portion of its mural photography during a migration completed in March 2026, when duplicate-detection software identified near-identical shots taken from slightly different angles and deleted what it calculated to be redundant copies — including, in several cases, the highest-resolution version of a given image rather than the lower-quality duplicate.

The Tenderloin Museum, on Turk Street, has faced a similar problem with its oral history project, which pairs audio recordings with portrait photographs of longtime Tenderloin residents. Multiple portraits were flagged and removed after the platform's algorithm matched them against stock photography databases, apparently because the lighting conditions and framing were close enough to commercial headshots to trigger false positives.

Community members who contributed those photographs have described the loss in stark terms. One woman who grew up in the Western Addition and donated a set of photographs documenting her family's return to the neighborhood after the redevelopment era of the 1960s said she had no backup copies. The images were the only photographic record of her grandmother's home on Fillmore Street before it was demolished.

What the Data Shows — and What Comes Next

The scale of the problem is difficult to quantify precisely because affected organizations are still auditing their collections. The Internet Archive, based in the Richmond District on Funston Avenue, has documented a broader national pattern: automated deduplication errors increased sharply after major cloud providers updated their image-matching algorithms in late 2025. The Archive has not released city-specific figures for San Francisco, but staff archivists have been in contact with at least six local nonprofits about recovery options.

Recovery is possible in some cases, but the window is closing. Most cloud platforms retain deleted files for between 30 and 90 days before permanent erasure. For organizations that completed migrations in March or April 2026, that window has already closed for a portion of their lost material.

The San Francisco Office of Civic Technology and Innovation has not yet announced a formal review of the grant program's data-handling requirements, though community advocates have requested one in writing. The next step for affected organizations is to file detailed loss reports with their platform providers and simultaneously contact the city's Department of Technology to request an audit of the migration process specifications — specifically, whether the contracts required disclosure of any automated content review applied to transferred data.

For residents who contributed personal photographs to these archives, the practical advice from digital preservation specialists is to check immediately whether the platform's deletion window remains open, and to request a full data export of whatever material survives before any further automated processes run.