Typhoon Noul hit China’s coast last weekend, and within hours the internet filled with corpses that never floated, crocodiles that were never released, and pigs that were never rescued ahead of people. None of it was real. All of it spread anyway.
Beijing’s response came fast: task forces, arrests, a joint notice from the Cyberspace Administration and the Ministry of Emergency Management ordering platforms to hunt down misleading disaster footage, recycled news reports, fabricated data, fake official announcements and disaster-related fraud. One case already made the rounds. A man in Hebei province generated a video claiming floodwaters had swallowed cars in a county that never flooded that way. He now faces prosecution under a campaign Beijing calls Clean Net 2026.
The fix looks reasonable on paper. It also skips the part where Beijing helped build the market for fake disaster footage in the first place.
The Trust Deficit Did This
Go back to the Liulan Dam. When its wall gave way near Nanning this month, local journalists were told to call the breach an “opening” instead of what it was. That account comes from a translated record of deleted commentary published by China Digital Times. Softer language, same collapsed wall, same flood downstream. At least 26 people died.
That’s not an isolated editorial choice. Henan officials went to prison in 2021 for hiding 139 flood deaths. Stack that pattern across two decades of provincial disaster reporting and the public learns something simple: don’t trust the first official number.
Fake AI videos don’t fill a vacuum of information. They fill a vacuum of credible information — one the state carved out itself. That gap is exactly where industrialized deepfake fraud tends to set up shop, whether the target is a flood zone or a bank account.
Fighting the Symptom, Not the Disease
Professor Chen Bin, writing in the Communist Party’s own Central Party School newspaper, put it plainly. Generating convincing fake disaster footage now takes nothing but a prompt. Professor Gao Fuping told the BBC the motives split three ways: some chase clout, some chase money, some just want to watch a region panic. None of that requires technical skill anymore. The barrier to fabricating a flood didn’t lower. It disappeared, part of a wider shift where deepfakes have become one of the defining risks of this stretch of the AI rollout.
Beijing’s answer is enforcement. A nationwide sweep instructs local cyberspace offices to flag AI-generated disaster scenes alongside impersonated officials and fake donation QR codes. Prison time, up to seven years, sits on the table for the worst offenders.
That catches the video creator. It does nothing for the credibility problem that made his video plausible in the first place — and enforcement without that fix is a governance problem dressed up as a technology problem. Shanghai and Washington-based lawyer Peter Pang wants platforms held liable too, arguing the people who “allow this to go out” share responsibility with the people who make it. Fair enough. But liability law doesn’t rebuild trust in a flood bulletin, and it lands in the same murky terrain defamation law is already struggling to define for synthetic content generally.
The Harder Problem Nobody’s Solving
Professor Xu Xiaoke’s proposed fix is the honest one. Build information infrastructure people actually believe before the water rises, not after. Official early-warning systems. Verified timestamps. Image provenance checks — the same instincts now showing up in advice ordinary readers get on catching a fake before it fools them. It’s a good idea that runs directly into the instinct — visible again in the Liulan euphemism, visible in the Henan cover-up — to manage the story before managing the disaster.
China has roughly 6,000 AI companies competing to make generation tools faster, cheaper, and more convincing, according to state wire service Xinhua. Detection is losing that race by design. Xu himself admits verification technology lags behind generation technology, and always will.
So the next flood arrives with two crises stacked on top of each other. One made of water. One made of pixels. Beijing can keep arresting the people who make the second one. Whether it can stop making the first one necessary is a different question — and it’s the one nobody in this crackdown is answering.
