by Tan Aik Keong (AK)
In July 2024, a case that came to be known online as the "Sister Hong" scandal set off a firestorm of public debate. At its centre was a man who, over a long period, posed as a woman on dating apps to lure men into sexual encounters, secretly filming the interactions and uploading them to paid group chats. The number of people implicated ran into the thousands by some accounts, and as the footage spread, it triggered widespread panic, online doxxing and a flood of rumour and speculation.
In the middle of that storm, a lot of ordinary users found themselves asking an uncomfortable question: could I — or someone I know — have ended up in one of these videos, or been wrongly identified through a doctored photo? That anxiety is what pushed some technologists to try using AI to help separate fact from fabrication.
A local AI tool steps in
One tool that got significant attention was HongjieGPT, a locally built AI platform. Reportedly developed in a short window by an AI team, it combines AWS Rekognition's facial-recognition system with a self-trained deepfake-detection model, built on the lightweight Streamlit framework and trained on roughly 1,691 publicly sourced images.
Users simply upload a photo, and the system checks it against a database of faces associated with the case, returning a similarity score and an authenticity assessment. If an image turns out to be fabricated, the deepfake-detection layer can flag signs of face-swapping or synthetic generation.
From what's been publicly tested, the system has correctly flagged several widely circulated photos as deepfakes, with fabrication probability scores above 80% in some cases — and it has also cleared public figures wrongly linked to the case, helping to shut down misinformation and online harassment aimed at them. Cases like this show AI's usefulness goes beyond business efficiency — it can play a real clarifying, protective role in messy, fast-moving social situations.
A public good, not just a corporate tool
AI isn't only for big companies and government agencies any more. Faced with online harassment and a fog of conflicting information, HongjieGPT shows what AI can do as a public good: a rapid-screening tool for separating fact from fiction, and a trust shield for protecting relationships and reputations.
Of course, any use of this kind of technology has to stay within the bounds of privacy, law and ethics. The platform behind it encourages users to query only publicly available, lawfully obtained images, and says it will keep strengthening its data-handling and access controls to prevent misuse. Technology can help identify — it shouldn't be used to pass judgment. Respecting the rights of both users and the people being checked is what will actually make tools like this sustainable.
Closing
The "Sister Hong" scandal started as gossip and scandal, but AI's involvement gave society a genuine tool for cutting through the fog of misinformation. Technology like this could eventually extend into fraud detection, digital reputation protection and missing-persons work — areas with real social value. Whether AI can truly become part of the social trust infrastructure will come down to how we define its boundaries, its use cases and its responsibilities. In the end, protecting relationships and reputations starts with being able to tell who's real — and in the information age, using AI to reduce collateral damage and replace suspicion with fact is technology doing what it should.
Part of the AK AI Corner column. Originally published in Oriental Daily (东方日报) on Jul 19, 2025.