中国爆发的 OpenClaw 热潮正考验北京对人工智慧的治理取向：在追求前沿创新与维护资料安全之间取得平衡。过去数周，agentic AI 迅速扩散，最高峰时深圳一间科技门市外曾出现「数百人」排队协助安装；但其作为外来技术的爆量采用引发关切，知情人士称中央政府于 2026-03-11（周三）警示国有企业与机关避免在办公电脑安装，以降低外部通讯与资料外泄、遭受外部攻击等风险。
监管上，中国未如欧盟采行全面性 AI 法，而是自 2022 年起以议题导向的临时措施处理特定风险（如演算法推荐、deepfake），并于「去年」率先强制标示 AI 生成内容；目前仍缺乏专门针对 AI agent 责任归属的规范。与此同时，隶属工业和资讯化部的 China Academy of Information and Communications Technology 表示将于 3 月下旬启动对 OpenClaw 等 AI agents 的可信度试点，并制定一系列使用标准；企业端如 Chery Automobile Co. Ltd 董事长 Yin Tongyue 亦要求员工暂缓安装，待完成训练计划后再导入。
数据面显示 AI 热潮与经济、就业及社会稳定的张力同步上升：北京设定 2030 年核心数位经济产业增加值占 GDP 比重由「去年」的 10.5% 提升至 12.5%，相当于 +2.0 个百分点（约 +19% 相对增幅）。然而在就业端，青年失业率已连续 6 个月徘徊在 15% 以上；Peking University 对「逾 100 万」则职缺的研究指出，最暴露于 AI 的部门（会计、编辑、程式设计等）已出现征才下滑。政府亦于 1 月表示正起草 AI 对就业影响的政策指引但未给出发布日期，个案如 Ryan Xie 与 Victor Chen 一方面以 Docker、Sandbox 等降低资安风险，一方面担忧 AI 驱动的大规模失业与潜在社会不安。
China’s OpenClaw frenzy is testing Beijing’s AI governance by forcing a tradeoff between cutting-edge innovation and data security. Over the past few weeks, agentic AI adoption surged; at its peak, “hundreds” queued outside a Shenzhen tech shop for installation help. The runaway uptake of a foreign tool drew official attention, and people familiar with the matter said the central government warned state-owned enterprises and agencies on Wednesday (2026-03-11) not to install it on office computers, citing risks tied to unusually broad data access and external communication that could expose systems to attack.
Regulatorily, China has avoided a sweeping AI law (unlike the EU), relying instead on ad-hoc rules since 2022 targeting specific problems such as algorithmic recommendations and deepfakes, and it became the first government “last year” to mandate labels for AI-generated content. There is still no rule specific to AI agents that resolves issues such as responsibility for an agent’s actions. The China Academy of Information and Communications Technology, affiliated with the Ministry of Industry and Information Technology, said it will begin a trustworthiness trial for agents like OpenClaw in late March and will develop a series of use standards; corporate responses include Chery Automobile Co. Ltd chairman Yin Tongyue telling staff to pause installation until a focused training program is in place.
Quantitative signals also underscore rising tension between AI-driven growth and social stability. Beijing targets raising the value added of core digital-economy industries to 12.5% of GDP by 2030 from 10.5% last year, a +2.0 percentage-point increase (about a 19% relative rise). Yet labor-market strain is evident: youth unemployment has hovered above 15% for the past six months, and a Peking University study of over 1 million job postings found recruitment declines in AI-exposed sectors such as accounting, editing, and programming. The Ministry of Human Resources said in January it was drafting guidance on AI’s job impact (no release date), while users like Ryan Xie and Victor Chen balance productivity gains with security workarounds (Docker, sandboxes) and fears of AI-driven mass unemployment and unrest.