Hong Kong's Duplicate Image Problem: The Key Decisions Coming This Autumn

Hong Kong's digital archival infrastructure is approaching a decision point. Across multiple government platforms — from the Land Registry's property documentation system in Queensway to the Hong Kong Public Records Office in the former Murray Building complex — tens of thousands of digitised files contain duplicate or misidentified images that have accumulated since the territory's mass digitisation push began in earnest after 2018. The question now is who fixes them, how fast, and who pays.

The problem is not unique to Hong Kong. Cities managing large-scale digitisation of legacy paper records routinely encounter image duplication rates between 8 and 15 percent, according to international digital preservation standards bodies. But in Hong Kong's case, the stakes are sharpened by two converging pressures: the government's stated ambition to position the city as a regional smart-city leader under the Greater Bay Area digital infrastructure framework, and growing scrutiny of public records integrity in the post-Article 23 environment, where the reliability of official documentation carries heightened legal significance.

What's Actually Broken — and Where

The core issue involves image deduplication — the technical process of identifying, flagging, and either merging or deleting redundant digital image files within document management systems. At the Companies Registry on Queensway, which processes hundreds of thousands of incorporation and amendment filings annually, duplicate scan files create indexing conflicts that slow retrieval times and, in some documented cases flagged internally since 2023, result in the wrong document image being served to a requesting party. The Lands Registry, headquartered at Queensway Government Offices, faces a similar challenge with scanned deed records dating back to the colonial era, where batch scanning produced duplicate page captures that were ingested without automated quality checks.

The government's Innovation, Technology and Industry Bureau, which oversees cross-agency digital standards, has been coordinating with the Office of the Government Chief Information Officer since late 2024 on a unified deduplication protocol. That protocol, initially expected to be finalised by the first quarter of 2026, has not yet been publicly released. The delay matters because individual bureaux have in the meantime been applying their own inconsistent remediation approaches, which risks compounding the problem rather than solving it.

Private-sector pressure is adding to the urgency. Law firms along Des Voeux Road Central, which rely on Lands Registry image pulls for conveyancing searches, have reported to the Law Society of Hong Kong that discrepant image results on the same title reference have caused delays in transaction completions. The Law Society has not issued a formal advisory, but the profession has been tracking the issue since at least mid-2025.

The Decisions That Will Define the Outcome

Three concrete choices now sit on the table for the relevant bureaux. First, whether to adopt a centralised deduplication engine managed by the OGCIO or to let each registry run its own remediation — a question with significant budget implications given that a centrally procured system covering the major registries would likely require a tender in the range of HK$40 million to HK$80 million over a three-year contract, based on comparable government IT tender awards published in the Government Gazette since 2022. Second, whether to retroactively audit and re-index all legacy digitised records back to 1997, or apply deduplication only to files ingested after a defined cut-off date, accepting that older records will carry residual error rates. Third, how to handle files where duplication has already propagated into downstream systems used by the Judiciary's eLitigation platform at the High Court on Queensway.

The OGCIO has scheduled a cross-bureau technical working group meeting for September 2026. That session is now widely regarded within government IT circles as the effective deadline for the centralisation-versus-devolution decision. If no unified framework emerges from September, individual agencies will almost certainly proceed independently, locking in incompatible approaches that a future consolidation effort would need to undo at considerable cost. Practitioners dealing with the registries daily would do well to document any retrieval anomalies now — those records will form the evidential basis for any post-remediation audit.