The problem became impossible to ignore by Tuesday. Administrators at the Infocomm Media Development Authority flagged a sharp spike in duplicate image submissions across several government-linked digital platforms, with the heaviest concentrations hitting the HDB resale portal and the National Heritage Board's heritage documentation system. The trigger, according to technical documentation reviewed by The Daily Singapore, was a combination of automated scraping tools and AI-generated photo variants that defeated older hash-based detection filters.
The timing is awkward. Singapore has spent the better part of three years positioning itself as a regional hub for responsible AI deployment, with the Smart Nation and Digital Government Office publishing its AI Verify framework and encouraging both public agencies and private sector players to adopt transparent, auditable digital systems. Running into a duplicate image crisis in mid-2026 — on platforms that millions of residents use to make major financial decisions — undercuts that narrative at a sensitive moment.
Where the Problem Is Hitting Hardest
The HDB resale portal, which processed more than 27,000 resale transactions in the whole of 2025, relies on property photo galleries to help buyers evaluate flats before committing to offers that frequently exceed S$700,000 in mature estates like Queenstown and Toa Payoh. When duplicate images — sometimes slightly recoloured or cropped variants of the same unit — are uploaded under different listing IDs, buyers risk making repeated enquiries on what is effectively a single property, distorting their sense of market availability.
The National Heritage Board's SG Pictorial Archive, which houses digitised photographs of landmarks from Chinatown's Smith Street to the former Kallang Airport, encountered a separate but related issue. Volunteer contributors uploading historical images through the public submission portal inadvertently re-submitted material that algorithmic tools had subtly altered, inflating the apparent size of the archive while diluting its accuracy. Board staff spent much of Wednesday manually reviewing flagged entries.
PropertyGuru, the Nasdaq-listed real estate platform headquartered in Mapletree Business City in Alexandra, confirmed it had activated its internal duplicate detection protocols after identifying an unusual volume of near-identical listing photographs across several Jurong East and Woodlands project pages. The company declined to give a specific count of affected listings, citing an ongoing internal review, but noted the issue was concentrated in the new launch resale category.
What Comes Next for Platform Operators
The IMDA has not issued a formal enforcement notice as of Saturday morning, but sources familiar with the agency's Digital Trust Centre indicated that guidance for platform operators on perceptual hashing — a technique that detects visually similar images even after minor edits — is expected before the end of July. Perceptual hashing goes a step beyond traditional MD5 or SHA-256 checksums, which can be defeated by changing a single pixel. Singapore's Government Technology Agency already uses a version of perceptual hashing in its CorpPass document verification pipeline, so the underlying technology exists within the public sector.
For ordinary residents, the practical advice is straightforward. Anyone using the HDB resale portal or PropertyGuru to shortlist flats should cross-reference listing addresses directly through the HDB Map Services portal at map.hdb.gov.sg before booking viewings. The map service pulls from a live transactional database that is harder to manipulate than individual listing photo galleries.
For cultural heritage projects, the National Heritage Board has asked volunteer contributors to hold new submissions until a revised upload interface — one that flags suspected duplicates in real time before finalising an entry — goes live. The board expects that interface to be ready by late August 2026.
The episode arrives as Singapore's digital platforms face wider questions about content integrity in an era of cheap, fast generative AI tools. Platforms that built their reputations on reliable, well-curated information now have to decide how aggressively they invest in detection infrastructure. The cost of inaction, measured in eroded user trust and misdirected housing searches, is climbing faster than the cost of better filters.