中国AI安全状况报告（2026）

July 2026 Authors This report was authored by Gabriel Wagner, Erik Lindblad, Kwan Yee Ng, and Brian Tse. Explore the interactive companion resources: https://aisafetychina.com/ Acknowledgements We are grateful to Charles Alaimo, Alan Chan, Jeffrey Ding, Karson Elmgren, Emmie Hine, KristyLoke, Zilan Qian, Huw Roberts, Matt Sheehan, Ziyue Wang, James Zhang, Jie Zhang, and JasonZhou for providing feedback and suggestions on this report. How to cite this report Gabriel Wagner, Erik Lindblad, Kwan Yee Ng, and Brian Tse, “State of AI Safety in China (2026),”Concordia AI, July 2026, https://aisafetychina.com/assets/State-of-AI-Safety-in-China-2026.pdf. Table of Contents Executive Summary Introduction and scope5 1Domestic Governance8 1.1High-level policy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .91.2Laws and regulations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .121.3Standards. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .18 2International Governance33 2.1Multilateral engagement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .332.2Bilateral engagement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .372.3International AI standards. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .402.4Non-official dialogues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .42 3Technical Safety Research48 3.1Methodology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .483.2Overall trends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .493.3Key research groups. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .493.4Specific research contributions. . . . . . . . . . . . . . . . . . . . . . . . . . . . .52 4Expert Views on AI Safety and Governance59 4.1Methodology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .594.2Safety risks from agentic AI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .604.3Open source AI governance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .624.4AI’s impact on cybersecurity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .644.5AI’s risks for biosecurity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .65 5Industry Governance68 5.1Industry-wide collective action. . . . . . . . . . . . . . . . . . . . . . . . . . . . .695.2Company-specific safety practices. . . . . . . . . . . . . . . . . . . . . . . . . . .71 Conclusion Notes81 Executive Summary As AI capabilities advance and agentic systems move from demonstrations into wide deployment,understanding how China approaches AI safety and governance remains essential to any serious effortat international coordination. Since 2023, Concordia AI’s annual reports on theState of AI Safety inChinahave tracked how China addresses risks from general-purpose AI. This year’s report provides updates from July 2025 to June 2026 across five domains: domestic gover-nance, international governance, technical safety research, expert views on AI safety and governance,and industry governance. Over the three years this report series has tracked, it has become increasingly clear that China’sapproach to AI safety and governance extends far beyond content control. Much of China’s earlyAI regulation in 2023 focused on what AIsays. Throughout 2024 and 2025, the rise of multimodalmodels moved the regulatory focus toward images, videos, and audio, prompting strict rules aroundlabeling AI-generated content. With the rise of autonomous agents and AI companion products overthe past year, the concern has expanded to what AIdoesand what itdoes to people and society, frommental-health impacts to labor-market disruptions. Domestic Governance •The 15th Five-Year Plan (2026–2030) cements “AI Plus” — the drive to diffuse AIacross sectors to spur economic growth — as China’s overarching AI policy, whilecontinuing to prioritize safety and risk management.Enshrined in March 2026, the plancarries forward President Xi Jinping’s April 2025 language on the importance of early risk warningand emergency response. It also highlights new concerns around AI’s impact on employment, withcalls for monitoring AI’s labor-market impact, strengthening reskilling programs and employmentsupport, and leveraging AI’s ability to create jobs. •The arrival of autonomous agents reoriented Chinese governance from controllingwhat AI says to controlling what it does.After the open source agent OpenClaw pro-liferated in early 2026, multiple cybersecurity authorities issued warnings, and in May 2026 theCyberspace Administration of China (CAC) and two other agencies issued dedicated guidance onagentic AI, devoting a full chapter to safety and proposing risk-based governance. Several majorAI standard-setting bodies are now drafting agent-related security standards. This marks a shift inChina’s framework from content control