SF Officials and Tech Experts Push Back Against AI Image Duplication Flooding City's Digital Records

San Francisco's Department of Technology is grappling with a surge in duplicate and AI-generated images jamming the city's digital record systems, a problem that archivists, urban planners, and open-data advocates say has gone from nuisance to genuine administrative crisis in the first half of 2026. The issue cuts across at least four city departments, including the Planning Department on South Van Ness Avenue, the Office of Digital Services, and the SF Public Library's digital preservation unit at the Civic Center branch.

The timing matters. The city spent the better part of 2024 and 2025 digitizing decades of building permit records, zoning maps, and neighborhood survey photographs under a $4.2 million modernization contract. That investment is now being undermined, according to city staff familiar with the project, by the automated ingestion pipelines that were supposed to speed up archiving but instead pulled in thousands of redundant image files — many of them near-identical AI-generated renderings submitted by developers seeking entitlements in neighborhoods like the Mission, SoMa, and the Tenderloin.

What Officials and Experts Are Warning

The SF Planning Department has flagged the problem internally since at least March 2026. Staff there have described a backlog in the environmental review queue partly attributable to duplicate visual materials — developer-submitted renderings that differ only in minor pixel-level ways but are treated as separate files by the city's document management system. The department processes thousands of project applications annually, and each extra file adds processing time and storage overhead to a system already under strain.

Urban data specialists at SPUR, the San Francisco planning advocacy organization headquartered on Sutter Street, have been tracking the broader trend of AI-generated content entering municipal workflows. Their researchers have argued that cities need explicit submission standards for digital images before automated review pipelines become standard — an argument that has gained traction after the duplicate-image problem surfaced. SPUR has not yet released a formal report on the San Francisco case, but staff there have discussed the issue publicly at community meetings this spring.

At the SF Public Library, the situation has a slightly different flavor. The library's digital collections team, which maintains the San Francisco Historical Photograph Collection — more than 200,000 images catalogued online — has seen an uptick in patron-submitted materials that include AI-generated or AI-enhanced images mislabeled as historical documents. Librarians are now manually flagging suspect submissions before they enter the permanent archive, a step that did not exist in the library's workflow before January 2026.

The Technical Problem Underneath the Policy Gap

The core issue, as technologists describe it, is that San Francisco's city systems were not built to detect perceptual duplicates — images that look identical to a human eye but carry different file hashes, meaning automated deduplication tools pass them through as distinct records. Consumer tools for catching this kind of duplication have existed for years, but government procurement cycles have not kept pace. The city's existing document management contract, administered through the Department of Technology on Seventh Street, does not include perceptual hashing as a required feature.

That gap is now the subject of a working group that convened for the first time in June 2026, drawing in staff from the Planning Department, the Office of Digital Services, and outside vendors. The group has no set reporting deadline yet, but participants have indicated a set of minimum technical standards for image submissions could be proposed to the City Administrator's Office before the end of the fiscal year in June 2027.

For residents and developers navigating the permit process right now, the practical advice from Planning Department staff is straightforward: submit the fewest number of image files necessary to document a project, avoid uploading multiple versions of the same rendering at different resolutions, and use PDF packages rather than loose image folders wherever the portal allows it. The department's public counter at 49 South Van Ness Avenue can advise on file format requirements before a formal application is submitted. The working group's preliminary findings are expected to be posted to the city's open-data portal, DataSF, once a draft is circulated internally.