China’s Ministry of State Security focuses on AI risks: political security is listed as the top challenge

(Deutsche Welle Chinese website) The safety and supervision of artificial intelligence have become a topic of international discussion in recent days. In a recent article published by Chen Yixin, China’s Minister of National Security, he listed the dangerous challenges that the development of artificial intelligence may bring. The first one listed is political security.

Chen Yixin wrote: "Various hostile forces and people with ulterior motives abuse generative artificial intelligence technologies such as deep fake audio and video, AI-generated graphics and text, and smart trolls to concoct political rumors at low cost and in batches, spread harmful information, incite oppositional sentiments, and carry out public opinion and cognitive warfare against me, directly threatening my political security, institutional security, and ideological security."

Rebecca Arcesati, head of the technology and innovation program at the Mercator Center for China Studies (MERICS), said in an interview with DW that this is not the first time that China’s Ministry of State Security has expressed concerns about AI risks. “At the World Artificial Intelligence Conference held in Shanghai in July this year, China’s top leader Xi Jinping publicly mentioned for the first time the risk of the government or all mankind losing control of AI systems.”

Alcesati said the issue should be viewed "from the perspective of an authoritarian one-party state obsessed with maintaining power." “That’s why the top concern cited by China’s Ministry of State Security is political security. Beijing is worried that hostile foreign powers will use AI systems to create instability within China, such as influencing public opinion and ultimately threatening the Chinese Communist Party’s ruling position. This is their top concern and explains why China is increasingly increasing its supervision of emerging technologies to ensure control.”

"WeChat Worm"

In the article, Chen Yixin mentioned the "Claude Mythos" and "GPT-5.5-Cyber" large models recently launched by American technology companies, saying that "some countries and organizations have the ability to quickly mine vulnerabilities in batches, automatically connect attack paths, and complete complex hacking tasks... which pose serious risks to my country's critical information infrastructure."

In terms of the risk of secret theft and leakage, Chen Yixin wrote: “Overseas espionage and intelligence agencies make extensive use of technologies such as intelligent web crawlers, intelligent data mining, and intelligent portrait analysis to collect and obtain sensitive information such as key national data, business secrets, and citizens’ personal privacy.”

Expert Alcesati mentioned the “WeChat worm” incident reported by the media recently. The New York Times reported on September 9 that thanks to advanced artificial intelligence models, a small research team at the Palo Alto, California-based security firm Calif recently took just over a week to build a hacking tool capable of wreaking havoc on WeChat, China's ubiquitous social media and messaging platform.

Alcesati analyzed this: "The underlying design of WeChat does retain internal interfaces specifically for monitoring by Chinese officials (especially the Cyberspace Administration of China). But if a foreign government or AI system suddenly acquires the ability to directly control Chinese WeChat user accounts, it will be a completely different scenario. As I said before, it will pose a huge potential threat to the political security of the CCP."

In response to a question about the safety of artificial intelligence on September 15, the Chinese Ministry of Foreign Affairs stated that China has always insisted on equal emphasis on development and security, attaches great importance to various endogenous and derived risks caused by artificial intelligence, is committed to building a solid bottom line for safety, and constantly improves laws and regulations, policy systems, application specifications, and ethical rules to prevent abuse and abuse, and ensure the safety, reliability, and controllability of artificial intelligence.