China adds new regulations covering AI risks after deepfake images and artificial earthquake images spread, and issues measures to force platforms to label them.
On November 13, 2025, at 2:36 a.m., Nikkei Asia reported that The Chinese government is moving to tighten regulations on fake images and videos created with generative AI technology. After discovering a large amount of misleading content on social media, including images claiming to be kidnapping victims or earthquake damage,
Amendments to the Internet Security Act will take effect in January. The bill was originally intended to protect the security of telecommunications infrastructure. But the latest amendments also include additional AI risk management and safety supervision measures. The amendments were approved by the Standing Committee of the National People's Congress on October 28.
The Chinese government is concerned that fake content could be used to criticize the state, particularly in cases where misinformation about crimes, accidents or disasters exaggerates the severity of the damage, potentially damaging China's image. There are also concerns that high-ranking officials could be slandered by false information and lose their positions.
China has had a generative AI service regulation law in effect since August 2566, stipulating that AI companies must use the technology in line with socialist values and must not create content that could cause social unrest or subvert the government.
DeepSeek and other generative AI models in China were trained under this law, leading to large-scale Chinese models often avoiding answering political questions. However, a recent surge in fake images and videos has prompted Chinese authorities to tighten controls on the technology.
Following the 6.8 magnitude earthquake in Tibet on January 7, a photo of a newborn baby covered in debris went viral on social media, with people expressing sympathy. However, it was later discovered that the image was AI-generated and unrelated to the actual event. The poster was detained for spreading false information.
Public security authorities have also intensified their crackdown. In late October, officials released samples of illegal fake content and announced they would prosecute other offenders.
Examples include a 28-year-old man in Zhejiang Province who posted photos of a girl online claiming to be his kidnapped daughter, and a 57-year-old woman in Shanxi Province who posted messages and photos of earthquake damage that never actually occurred.
The number of generative AI users in China reached 515 million as of June, doubling in just six months. Its adoption has exploded after DeepSeek gained traction for its powerful capabilities and low cost.
Browsing the Chinese internet, you'll find a plethora of AI-powered tutorial videos, including face and voice-changing programs. Many users post provocative images to drive traffic, boost sales, or generate advertising revenue.
Internet regulators linked to the Chinese government and the Communist Party have begun issuing stricter guidelines to local governments and major social media platforms. In a crackdown on AI services that can easily create pornographic images or alter voices and faces, the agency ordered the revisions of 3,500 apps and removed 960,000 pieces of content from the April-June quarter.
In September, the agency announced new requirements that will require AI providers to label user-generated images or videos to identify them as AI-generated content.
