OMM Redaksiyası
Warnings from researchers at leading US artificial intelligence company Anthropic that increasingly powerful AI models could escape human control have drawn attention in China as well. The United States and China are considered the leading countries in the development of advanced AI and the global spread of the technology.
Both countries share concerns about controlling artificial intelligence, but their approaches to safety measures differ. Beijing relies more heavily on state oversight, while discussions in the United States largely focus on control mechanisms implemented at the company level.
Chinese policymakers view artificial intelligence as a powerful but manageable technology. In their view, its risks can be limited through technical standards, regulation and state supervision.
Brian Tse, founder and chief executive of Concordia AI, said Chinese and US experts broadly agree on the risks posed by artificial intelligence. According to him, the difference lies in how those risks are framed and prioritized.
In a policy document issued in May by Chinese cyberspace, economic planning and industrial authorities, China identified “loss of operational control” over AI agents as a security risk. Unlike traditional chatbots, these agents can plan and carry out multistep tasks with greater autonomy.
The document requires developers to strengthen their ability to detect, intervene in, block and recover from improper or uncontrolled behavior by AI agents. It also recommends measures against risks such as data poisoning, algorithm manipulation and system vulnerabilities.
Under Chinese regulations, an artificial intelligence system must be transparent, users must be informed about agents’ autonomous decisions, and final decision-making authority must remain with humans.
Chen Yixin, China’s minister of state security, wrote in a government-affiliated publication on Sept. 13 that advanced models such as Anthropic’s “Mythos” and OpenAI’s “GPT-5.5-Cyber” could pose serious threats to China’s critical information infrastructure. He called for a comprehensive strengthening of AI security.
China first explicitly mentioned a future scenario involving loss of control in its artificial intelligence safety framework released in September 2024. The document said the possibility could not be ruled out that AI might independently acquire external resources, replicate itself, develop consciousness and seek external power.
The framework, updated in September 2025, provided further detail on the scenario. The new document emphasized that artificial intelligence could experience a sudden and unexpected “leap” in intelligence before acquiring resources, replicating itself or seeking power.
China’s regulatory approach differs from calls by some Western AI safety circles to slow or halt the development of the most powerful models until safety mechanisms are created. Instead, Beijing promotes the integration of artificial intelligence across all industries, although it can delay implementation when it believes governance is lagging behind technological development.
