Singapore Punishes the Tool, Not the Lie

A man in Singapore faces ten years in prison for an image that never happened — a crocodile generated by artificial intelligence in a reservoir where no crocodile existed, where the panic lasted only as long as the image circulated before deletion.

The sentence is not proportional to the injury. It is proportional to the law's terror of the medium itself.

This pattern runs deep. In the 1980s and 1990s, Malaysian authorities prosecuted citizens under supernatural hoax statutes for circulating photographs that supposedly showed ghosts — photographs that predated digital cameras by years, when people had been sharing phantasmagoric images since the earliest optical illusions. Courts convicted under acts written for pamphlets and oral claims, treating the photograph as the crime rather than the lie.

Singapore is doing this now with generative AI. The charge is not that the image caused measurable panic, though it did, nor that specific people were harmed by a false belief that a crocodile lurked waiting, nor even that the creator intended to deceive. The charge is that someone made a synthetic image that looked like reality. And the law knows how to punish the making of synthetic images the way 1980s Malaysia knew how to punish the owning of ghost photographs, by treating the tool as the transgression. Each new medium arrives into a legal system built for the last one. Photography was prosecuted under rules for paintings, video under rules for photography. Now generative images under rules for deepfakes, each layer assuming the previous threat was the real one. The actual variable. Whether anyone died, whether the panic persisted, whether the false belief changed behavior. Remains invisible to the sentence.

What changes this time is speed. A crocodile that never existed can be unseen in seconds, while a ghost photograph required physical prints and door-to-door distribution. The interval between creation and deletion now happens at digital velocity. Means the law is finally catching something that self-corrects faster than it spreads. Whether prosecutors recognize this before the next conviction, whether they separate the genuine threat from the medium's mere capacity to threaten. That becomes the only question that matters.