September 16, 2026, 06:38 GMT | Insight
China is stepping up its push to ensure artificial intelligence develops in a safe, reliable and controllable way, with a senior cyberspace regulator calling for agile oversight, stronger built-in safeguards and closer coordination across government and industry. Speaking at a forum in Jinan, Wang Jingtao, deputy director of the
Cyberspace Administration of China, outlined initiatives to address the intrinsic and downstream risks AI can create, including social, environmental and ethical harms.
China is stepping up its push to ensure artificial intelligence develops in a safe, reliable and controllable way, with a senior cyberspace regulator calling for agile oversight, stronger built-in safeguards and closer coordination across government and industry.
Speaking at a forum* in Jinan, Wang Jingtao, deputy director of the Cyberspace Administration of China, outlined initiatives to address the intrinsic and downstream risks AI can create, including social, environmental and ethical harms.
The so-called “intrinsic” AI risks arise from flaws or vulnerabilities in models, algorithms, training data, computing infrastructure and operating environments. The emphasis signals a broader shift in China’s AI governance from content and data controls toward a more comprehensive framework addressing risks embedded in the technology itself.
Wang’s remarks come as AI leaders in the international community debate slowing frontier-model advances to give safety research, testing and regulation more time.
Anthropic Chief Executive Officer Dario Amodei has called for such a pause and said the company will allow independent evaluators to monitor its model training and safety practices.
OpenAI’s Sam Altman backed similar arrangements, while Elon Musk wrote on X: “Dario is right.”
Wang cited President Xi Jinping’s call at the 2026 World Artificial Intelligence Conference to address such risks as part of efforts to build a fair and equitable global AI governance system (see
here). Xi urged the development of legal frameworks, technical monitoring, risk-warning and emergency-response systems to prevent misuse and ensure AI remains under human control.
Wang said regulators should follow principles of “inclusive prudence” and “agile governance,” continuously improving AI safety oversight through targeted and flexible tools that contain risks without holding back technological and industrial development.
He said China should actively explore new regulatory approaches, including sandbox supervision, as AI products and applications evolve. Such mechanisms can allow companies to test new technologies in controlled conditions, giving regulators visibility over emerging risks while preserving room for experimentation.
The balancing act reflects China’s push to use AI to support growth, industrial upgrading and public services while guarding against threats to social order, public interests and national security.
Wang also urged key industries to set application-specific safety rules based on their business needs. He said stronger industry standards should ensure AI adoption supports sustainable growth without compromising security.
At the same forum, the China Electronics Standardization Institute, which is affiliated with the Ministry of Industry and Information Technology, released a series of AI-application security guidelines. The package includes a general document intended to provide cross-industry security guidance for organizations using AI. It also contains more tailored guidance for education, health care, broadcasting, television and online audiovisual services.
— Stronger built-in safeguards —
While flagging “intrinsic” AI risks, Wang cited some tests that have already raised warning signs, including models breaking through sandbox isolation and operating beyond their intended boundaries as large language models and AI agents advance rapidly.
To address those issues, Wang called for moving AI security governance upstream by embedding safety standards and ethical safeguards throughout the development, training and iteration of foundation models. He urged a continuous, intelligent risk-management system spanning detection, early warning, optimization and emergency response.
That approach could require closer coordination between AI developers and cybersecurity companies. Wang urged greater investment and technological breakthroughs in testing methods, mechanisms and tools that can improve the ability to discover, judge and validate safety risks.
Wang also called for closer oversight of emerging AI businesses, including the evolving use of AI agents. He said safeguards should set clear behavioral boundaries, restrict permissions and block risky actions, embedding security into technical workflows to prevent risks at their source.
— Joint efforts —
Wang called for a governance structure led and guided by the government but supported by corporate responsibility, research institutions and broader social participation.
Government departments should strengthen risk-information sharing and coordinated responses, while companies must take primary responsibility for safety by embedding safety principles and standards throughout technological innovation and business development, Wang said.
“Companies should proactively identify and promptly address potential security risks, serving as the first line of defense,” he said.
Wang also called for industry, academia and research institutions to work together to strengthen risk awareness and prevention capabilities, advance standards and technical cooperation, and address threats including technology misuse and loss of control.
*"AI Security Governance Forumm" part of China Cybersecurity Week, hosted by the Cyberspace Administration of China and Cyber Security Association of China, Jinan, Shandong Province, Sept.15, 2026
- Analysis by Olivia Wang.
Please e-mail editors@mlex.com to contact the editorial staff regarding this story, or to submit the names of lawyers and advisers.
Tags
Sections:
Data Privacy & Security, Artificial Intelligence
Industries:
Automotive, Banking & Finance, Consumer Products, Computing & Information Technology, Educational Services, Entertainment & Arts, Health Care, Insurance, Manufacturing, Marketing & Advertising, Media & Telecommunications, Retail & Wholesale Trade, Transportation & Warehousing, Computer Networks, Social Media, Digital Economy
Geographies:
Asia, China, Eastern Asia
Topics:
Cybersecurity