Twenty-five cases.

On July 13, the Cybersecurity Bureau of the Ministry of Public Security dropped 25 typical cases of flood and disaster-related rumors all at once. Multiple cases explicitly involved using AI tools to generate fake videos — this isn't the first crackdown on AI rumors, but releasing this many cases in one batch is more aggressive than expected.

A netizen named Li in Qingyuan, Guangdong used AI to generate a fake video of "a distillery submerged by floodwaters," which spread and caused negative social impact. A netizen named Chen in Jiangsu made an AI video of "a reservoir dam breach" and forwarded it in multiple groups. Others used AI to generate fake footage of "rescue teams ignoring people" or "aid supplies being diverted," deliberately制造 social panic.

These cases share a common trait: extremely low production cost, extremely fast spread, enormous harm. Before, spreading rumors required some Photoshop skills, some video editing ability. Now? An AI video generator, type in some text, and in seconds you have a video indistinguishable from reality. The barrier to entry is on the floor.

When the Cost of Falsification Approaches Zero

The普及 of AI content generation technology presents content governance with a brand new problem: the marginal cost of造假 is nearly zero.

How did rumor-mongering used to work? You had an idea, then found素材, Photoshopped images, edited video, wrote copy — at minimum tens of minutes to hours, and you needed some technical skills. And spreading it? You needed channels, followers, groups — otherwise nobody saw it.

Now? The process is: open an AI tool, type a one-sentence description, wait a few seconds, the video is generated. Then one-click share to a few groups, a few platforms — done. The whole thing takes less than 5 minutes, no professional skills required, anyone can do it.

This leads to one outcome: rumors explode in volume, and governance can't keep up. You just spot one fake video, and ten new ones have already been generated. You just delete content from one platform, and it's already spread to five others. The traditional "discover-verify-delete" governance model is increasingly outmatched in the AI era.

What makes it even harder is the "realism" problem. Early AI-generated videos — look closely and you could spot the tells: weird fingers, facial twitching, blurry edges. But by 2026, AI video generation quality has reached an extremely high level. Ordinary people can't tell the difference with the naked eye. You see a "dam breach" video in a group chat — how do you judge whether it's real or AI-generated? Most people can't. They just instinctively forward it — "what if it's real? Forward it to warn people."

It's that "what if it's real?" mindset that makes AI rumors spread orders of magnitude faster than traditional rumors.

A New Phase of Governance: From "Post-Hoc Deletion" to "Proactive Prevention"

By releasing 25 cases in one batch, the Ministry of Public Security is sending a clear signal: AI rumor governance has moved from the "discussion phase" to the "law enforcement phase".

Before, when people discussed AI misinformation, it was mostly at the academic, technical, ethical level — "will AI-generated content cause problems?" "how should we respond?" No more discussion now. Straight to cases, straight to penalties, straight to enforcement.

But is post-hoc punishment enough? Probably not. AI rumors are produced too fast — no matter how fast you penalize them, you can't keep up with generation speed. And many of the rumor-mongers are just regular netizens, not professionals — they might have just done it on a whim, for fun, for a bit of attention. Do you really arrest all of them? Neither realistic nor necessary.

So what matters more is proactive prevention. For example: AI-generated content must carry watermarks, must have clear labeling. Platforms need to建立 AI content detection mechanisms — suspected AI-generated fake content gets throttled first, then verified. AI tool vendors bear some responsibility; they can't let their tools be abused. And public media literacy needs to improve — people need to learn to maintain skepticism toward information in the AI era.

This is a systems problem, requiring joint effort from regulators, platforms, AI vendors, and users. No single party can solve it alone.

But regardless, the Ministry's batch release of cases is a positive signal — at least it shows that governance is taking this seriously and starting to act. The AI rumor problem: better to govern early than late, better to be proactive than reactive.

After all, when true and false become harder and harder to tell apart, trust becomes the scarcest resource of our time.

See you tomorrow.