How Singapore's War on Duplicate Images Went From Bureaucratic Afterthought to National Priority

Singapore's public sector has quietly been sitting on a problem for years: thousands of duplicate images lodged inside government portals, national identity databases, and municipal record systems, consuming storage, slowing verification workflows, and — in the more serious cases — generating compliance headaches under the Personal Data Protection Act. The push to systematically address the issue, now formalised under the Government Technology Agency's (GovTech) broader Smart Nation data quality drive, did not arrive suddenly. It built up across nearly a decade of digital migration decisions that prioritised speed over hygiene.

The timing matters because Singapore is no longer treating data infrastructure as a back-office concern. The country's ambitions as a regional AI hub — anchored physically around one-north in Buona Vista and the expanding Punggol Digital District — depend on clean, non-redundant datasets feeding machine-learning pipelines. Duplicate imagery inside government systems is not merely an annoyance. When facial recognition tools or document-verification algorithms train on duplicated inputs, accuracy degrades. For a city-state that processed more than 350,000 MyInfo transactions per month as of 2024, even marginal error rates compound quickly.

How the Duplication Problem Accumulated

The roots go back to the mid-2010s, when multiple agencies digitised their paper records in parallel rather than in sequence. The Housing and Development Board scanned tenancy and ownership documents. The Immigration and Checkpoints Authority migrated identity photographs. The Central Provident Fund Board digitised member records. Each agency built its own repository. Integration came later — and imperfectly. When the Singpass app unified citizen-facing digital services from 2018 onward, back-end image assets from legacy systems were often carried across wholesale rather than deduplicated first.

Within the private sector, the picture was no cleaner. Property listing platforms serving the Toa Payoh and Tampines resale corridors — among the busiest HDB transaction zones — routinely hosted the same unit photographs uploaded by multiple agents. Platforms operating under the Council for Estate Agencies' licensing framework acknowledged the issue in industry consultations held in 2023, though no sector-wide deduplication standard was mandated at that point.

The inflection came in early 2025, when GovTech published its Digital Government Blueprint refresh. The updated framework explicitly named data deduplication as a prerequisite for responsible AI deployment within public agencies, setting a target of reducing redundant digital assets in core government repositories by 40 percent before the end of 2026.

What the Clean-Up Actually Involves

Deduplication at this scale is not simply a matter of deleting matching files. Images taken under different lighting conditions, compressed at different ratios, or cropped differently by different operators can represent the same source photograph without being byte-for-byte identical. Perceptual hashing — a technique that generates a fingerprint based on visual content rather than raw file data — has become the preferred tool for agencies working through backlogs. GovTech's Government Commercial Cloud infrastructure, hosted in part at Tanjong Pagar's data centre cluster, now runs automated deduplication sweeps as part of routine data governance cycles.

The Infocomm Media Development Authority has separately been working with media organisations and content platforms to establish cleaner ingestion protocols, reducing the volume of duplicates entering archival systems at the point of upload rather than retrospectively. Pilot programmes ran through the second half of 2025 with at least three local news publishers, according to published IMDA programme documentation.

For ordinary Singaporeans, the practical dividend will arrive gradually. Faster document verification on Singpass, fewer redundant images crowding property search results, and more reliable government digital services are the near-term payoffs. For businesses handling customer image data — insurers in the Raffles Place financial district, for instance, or logistics firms in Jurong — the sharper data quality standards being modelled in the public sector are already being cited in PDPC advisory guidelines as best practice for private operators.

The 40 percent reduction target carries a hard deadline of December 31, 2026. Agencies that fall short face mandatory audit reviews in the first quarter of 2027. Whether the timetable holds is the open question — but the frameworks, the tools, and the political will to enforce them are further advanced now than at any point in Singapore's digital governance history.