Explained: How China Is Tackling The Threat Of AI Escaping Human Control
US and China have been increasingly at loggerheads over each other's AI policies and industry practices, with these issues slated to feature prominently in bilateral talks later this month.
China is taking steps to mitigate the potential risks of artificial intelligence (AI) systems escaping human control and posing existential threats, according to warnings from leading US AI developers. Both China and the US are among the world's primary drivers of advanced AI technology, but their policies and practices often put them at odds.
Despite this competitive relationship, Beijing acknowledges the seriousness of the possibility of advanced AI systems functioning beyond human oversight and is preparing accordingly. Chinese state security minister Chen Yixin expressed concerns in a government outlet about US models such as Anthropic's Mythos and OpenAI's GPT-5.5-Cyber posing risks to China's critical information infrastructure.
He called for a comprehensive strengthening of AI security. Chinese developers have embraced open-weight models, partly because cybersecurity teams can review, alter, and employ them for defensive purposes. Hugging Face, a model repository platform, utilized an open-weight model developed by China's Z.AI after more closely monitored US models proved insufficient for forensic analysis.
However, experts caution that open-weight models can also be modified and distributed with minimal oversight, posing their own risks. In September 2024, China included a future loss-of-control scenario in its AI safety framework, developed under the guidance of the Cyberspace Administration of China (CAC). The framework has been expanded since, emphasizing the need for "trusted application, preventing loss of control."
In July, President Xi Jinping emphasized the importance of human control over AI, while Sun Xiaobo, a senior foreign ministry official responsible for AI affairs, announced China's acceleration of research into broader AI legislation. China has also issued guidelines specifically addressing AI agents, requiring developers to strengthen their capacity to detect, intervene, block, and recover from improper agent behavior.
These guidelines identify data poisoning, algorithm manipulation, system vulnerabilities, and "operational loss of control" as significant security risks, while retaining decision-making authority with users.
Written by urgent.news from Dawn's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
This story
This is one outlet's version. Read the fullest account.
