It seems that the start of a Chinese video AI prevents sensitive photos politically

China -based emerging, Sand AIShe released a licensed model of video artificial intelligence, which has been praised by entrepreneurs such as Microsoft Research Assia SUGENTING KAI-FU Lee. But Sand AI seems to be controlling images that may raise the anger of the Chinese organizers from the host version of the model, according to the Techcrunch test.

Earlier this week, Sand Ai announced Maggie -1It is a model that creates videos through “self -compression” to predict the tire sequences. The company claims that the model can generate high -quality shots that can be controlled more accurately embodies the competing open models.

👀 You will not believe that this is created from artificial intelligence
🧠 You will not believe it is open source
🎬 You will not believe it is free

Magi-1 video model only degrading commercial video tools

Details and examples below: 👇 pic.twitter.com/zlxrecweqh

Farhan (MHDFARAN) April 21, 2025

Magi-1 is not very practical so that it cannot be run on most consumer devices. It is 24 billion, and it requires between four and eight NVIDIA H100 graphics processing units to operate it. (Teachers are the models of the internal variables that you use to create predictions.) For many users-this reporter-Sand AI platform is the only place they can test Magi-1.

The platform needs a “quick” image to start generating the video. Not all claims are allowed, Techcrunch has been discovered quickly. Sand AI prevents downloading images from Xi Jinping, Square Tiananmen and Tank Man, Taiwanese flag, and the emblem that supports Hong Kong editing. The liquidation appears to occur at the image level; Naming photo files did not avoid the ban.

SAND AI is not the only Chinese startup that prevents sensitive photo downloads to its video generation tool. Hailuo Ai, the Shailuo AI, the Shanghai -based Minimax, prevent Xi Jinping images as well. But it seems that Sand AI’s liquidation is especially aggressive; Hauiluo allows Tiananmen Square photos.

As shown in a piece of January, models in China must follow up on strict information control elements. Law 2023 prohibits examples from generating content “harmful to the unity of the country and social harmony” – that is, opposing the historical and political accounts of the government. For compliance, Chinese startups often monitor their models, either by rapidly level filters.

It is interesting, while Chinese models tend to prevent political speech, they often have fewer filters than their American permitted content. 404 mentioned recently A number of video generators released by Chinese companies lack the basic handrails that prevent people from generating non -high nudity.