Singapore's Fight Against Fake Images: What Officials, Experts and Key Figures Are Saying

The problem has a deceptively simple name. Duplicate image replacement — the process of identifying, flagging and substituting manipulated or recycled photographs circulating online — has quietly become one of the sharper pressure points in Singapore's broader campaign against digital misinformation. Regulators are no longer treating it as a technical footnote.

The urgency is real. Singapore's Protection from Online Falsehoods and Manipulation Act, better known as POFMA, has been in force since 2019, but its original architecture focused heavily on text-based falsehoods. Images — cloned, cropped, recoloured or generated wholesale by artificial intelligence — represent a gap that both the Infocomm Media Development Authority and the Ministry of Digital Development and Information have signalled they want to close. The government's Digital Connectivity Blueprint, published in 2023, flagged AI-integrity infrastructure as a national priority, and that framing has only grown more pointed heading into 2026.

At the National University of Singapore's Institute of Data Science on Kent Ridge Drive, researchers have been working on perceptual hashing techniques — a form of digital fingerprinting that detects near-identical images even after minor edits like brightness tweaks or aspect-ratio changes. The work feeds into a broader ecosystem. Singapore Press Holdings' fact-checking unit, which operates under the SPH Media Trust on Toa Payoh North, has cited image duplication as among the most common vectors for viral misinformation on local WhatsApp channels, particularly around election cycles and public health announcements.

Where the Debate Sits Right Now

The core disagreement among experts is not whether duplicate images should be replaced, but who holds the authority to make that call — and how fast. Platform speed is the crux of it. An image can reach tens of thousands of Singaporean users within thirty minutes of posting; formal government correction notices under POFMA typically take hours longer than that. Civil society groups, including the Singapore chapter of the Alliance for Independent Journalism, have argued that automated replacement tools risk over-correction, pulling legitimate news photographs that share visual similarities with flagged content.

The IMDA has, in various industry consultation rounds, pushed back against that characterisation. The agency's position — drawn from public consultation documents released in late 2024 — is that human-in-the-loop review remains mandatory before any substitution is executed on a licensed platform. That human checkpoint is where technologists and regulators have found the most friction. Smaller newsrooms and community portals, particularly those serving Singapore's Malay and Tamil-language audiences, have raised concerns about whether they have the staffing to meet that standard in real time.

Meta, which operates Facebook and Instagram from its Asia-Pacific regional hub at One-North in Buona Vista, has engaged with IMDA on image-integrity protocols. Google's Singapore office, based in Pasir Panjang, has similarly participated in working groups. Neither company has publicly disclosed the specific technical benchmarks agreed upon with local regulators.

The Practical Stakes

The numbers give the debate its weight. Singapore ranks among the top ten countries globally for smartphone penetration, with more than 88 percent of residents owning one, according to the Infocomm Media Development Authority's 2024 annual survey. That density means image-based misinformation scales extraordinarily quickly across the island. During last year's municipal by-election in Mountbatten, at least three recycled images from unrelated overseas protests were circulated with local captions before fact-checkers intervened.

Researchers at Singapore Management University's School of Computing and Information Systems, located on Stamford Road in the Civic District, have proposed a tiered replacement framework: low-confidence matches get a visible watermark and a link to the original source, while high-confidence duplicates are substituted outright with a correction label. That model has attracted interest from the government's Smart Nation and Digital Government Office, though no formal adoption timeline has been announced.

For ordinary users, the practical advice from digital literacy advocates at the National Library Board — which runs its S.U.R.E. programme across all 28 public library branches — remains consistent: reverse-image search any photograph that prompts a strong emotional reaction before sharing it. On the regulatory side, the next twelve months will be telling. IMDA's scheduled review of the POFMA correction framework is due before the end of 2026, and image-integrity provisions are expected to feature prominently in whatever amendments emerge.