Cities Combat AI Fakes Flooding Digital Archives Worldwide

Digital archivists and city heritage officials are sounding the alarm. Duplicate and AI-generated images are infiltrating public records, municipal planning databases and cultural archives at a pace that administrators say they were not equipped to handle, and the damage to historical accuracy, they warn, is compounding by the month.

The problem has sharpened this summer, set against a broader backdrop of information disorder. As Tehran crowds gathered for a state funeral broadcast across hundreds of platforms, observers noted how the same photographs, some authentic, some algorithmically altered, circulated simultaneously in news feeds, government portals and Wikipedia-adjacent repositories. That episode, archivists say, crystallized a threat that has been building since at least 2023, when generative image tools became widely accessible without licensing restrictions.

What Heritage Bodies and City Officials Are Warning

In Berlin, the Landesarchiv, the state's official public records authority, acknowledged earlier this year that duplicate-image contamination had become a standing item on its quarterly audit schedule. The archive, which holds photographic documentation stretching back to the late 19th century, began piloting a hash-based deduplication protocol in January 2026 across its digitised collections, which run to more than 4.2 million indexed files. Archivists there have not publicly attributed a specific failure rate to synthetic images, but the decision to introduce automated screening reflects an institutional judgment that manual review alone is no longer sufficient.

In Cairo, the Egyptian Heritage Rescue Foundation has been working alongside the Supreme Council of Antiquities since the announcement last month of a newly uncovered Byzantine city in the western desert. That find, significant in its own right, immediately raised the issue of image provenance: how do official bodies ensure that photographs submitted by contracted field teams are originals, unmodified and free of duplicates pulled from earlier digs? The foundation has been advocating for a centralized metadata registry, a proposal it first circulated in draft form to regional partners in March 2026.

Urban planning departments face a parallel version of the same crisis. In London, the Greater London Authority's planning portal, which processes hundreds of developer submissions monthly, updated its image submission guidelines in April 2026 to require cryptographic watermarking on all site photographs. The policy change, noted in the GLA's April planning bulletin, followed an internal review that found a small but non-trivial number of planning applications in the 2024-25 cycle had contained photographs that appeared in more than one submission, sometimes for sites kilometres apart.

Expert Consensus and Practical Pressure

Specialists in digital forensics point to a narrowing window for intervention. Hany Farid, a professor at the University of California Berkeley whose published research covers image authentication, has written extensively about the asymmetry between tools that generate synthetic images and tools that detect them, a gap that continues to widen. His peer-reviewed work, published across several journals since 2020, has been cited by policy bodies in the European Union as foundational to the EU's AI Act provisions on synthetic media, which came into force in August 2024.

The International Council of Museums, headquartered in Paris, issued a technical advisory in May 2026 recommending that member institutions adopt Content Credentials, an open standard developed by the Coalition for Content Provenance and Authenticity, as a baseline requirement for any digitisation project receiving public funding. The advisory stopped short of mandating enforcement mechanisms, a point that critics inside the museum sector say leaves the guidance largely toothless without accompanying budget commitments.

The cost of retrofitting existing archives is not trivial. Estimates circulating among European heritage consortia suggest that a mid-sized municipal archive holding one million digitised assets could face implementation costs of between €80,000 and €150,000 to achieve full Content Credentials compliance, depending on existing infrastructure. For smaller institutions, that figure represents a significant share of an annual digitisation budget.

For anyone maintaining a public-facing digital archive, whether a city planning office, a university library or a community heritage project, archivists recommend three immediate steps: run existing collections through an open-source perceptual hash checker such as pHash, establish a clear submission policy requiring original file metadata, and designate a named staff member responsible for provenance verification on every new intake. The tools exist. The question officials and experts keep returning to is whether the institutional will, and the budget, will follow before the contamination becomes irreversible.