How Hong Kong's War on Duplicate Images Reached a Breaking Point

Hong Kong's public sector and media organisations are sitting on a sprawling, largely unaudited backlog of duplicate digital images — files that have quietly multiplied across servers, archival databases, and content management systems for more than fifteen years, costing storage budget, slowing publishing workflows, and, in several documented cases, generating legal confusion over which version of a photograph constitutes the official record.

The problem did not appear overnight. It is the accumulated result of decisions made — and deferred — across three distinct phases of Hong Kong's digital transition, each of which added another layer of redundancy without a coordinated plan to address the one before it.

Three Phases That Built the Problem

The first phase ran roughly from 2005 to 2012, when newsrooms along Wanchai's Lockhart Road and government communications offices in Central migrated from film and early low-resolution digital capture to high-megapixel workflows. Cameras improved faster than storage protocols. Photographers uploaded multiple versions of the same shot — RAW, JPEG, thumbnail — to shared network drives with no enforced naming convention. The Hong Kong Government Records Service, headquartered in the former Victoria Public Mortuary building on Hing Wong Terrace, received batches of image files from departments that had already duplicated them internally two or three times.

The second phase, roughly 2013 to 2019, compounded matters. Cloud migration became policy across large organisations, including RTHK and several of the major English-language publishers operating out of Causeway Bay. Files were uploaded to cloud buckets while the originals remained on local servers — a belt-and-suspenders instinct that made practical sense at the time but doubled storage overhead. Industry surveys conducted by the Hong Kong chapter of the Digital Asset Management Society in 2018 found that duplication rates across surveyed media archives exceeded 40 percent of total stored image volume, a figure that placed Hong Kong broadly in line with comparable markets in London and Toronto.

The third phase is the one that has forced the issue into the open. Post-2020 governance changes under the National Security Law introduced stricter requirements around the integrity and traceability of official records. Article 23 legislation, enacted in March 2024, further sharpened accountability expectations for organisations handling public-interest content. When a government document or a court filing references a specific photograph as evidence or official record, the existence of five visually identical but technically distinct file versions creates an audit trail problem that lawyers and records managers have been flagging with increasing urgency throughout 2025 and into this year.

What the Industry Has Done — and Hasn't

The Hong Kong Productivity Council, based in Kowloon Tong, has been piloting perceptual hashing tools — software that assigns a fingerprint to image content rather than file metadata — with a small cohort of public-sector clients since January 2025. The technology can flag near-duplicate images with a reported accuracy rate that vendors place above 95 percent for standard photography, though the council has not published independent verification of that figure.

Several commercial photo libraries operating out of Wong Chuk Hang's converted industrial studios have adopted similar deduplication pipelines on their own initiative, driven less by regulation than by the practical economics of storage costs. Server space in Hong Kong's Tier 3 and Tier 4 data centre facilities — concentrated in Tseung Kwan O — remains among the most expensive in Asia, with co-location pricing benchmarks roughly 30 to 50 percent higher than comparable facilities in Singapore, according to property consultancy data circulated at the DataCloud Asia conference in late 2024.

For organisations that have not yet addressed the backlog, the practical priority now is establishing a baseline audit before any deduplication tool is deployed. Records managers recommend cataloguing file origins first — distinguishing between duplicates created through workflow error and those created intentionally as backup copies — because the legal status of each category differs under the current records ordinance framework. The Government Records Service has published updated guidance on digital image provenance, accessible through its portal, which sets out the documentation standard that public bodies are expected to meet by the end of the 2026-27 financial year.