人工智能安全的国际合作新视角
中国在人工智能治理与安全议题上展现了开放的国际合作姿态。近年来，中国陆续发布了多项政策文件和治理倡议，强调在推动AI技术发展的同时确保安全性。此外，中国的研究机构、科学家及政策学者也积极投身于全球AI安全议题的探讨。
China has consistently shown an open stance toward global cooperation on AI governance and security. In recent years, a series of policy documents and governance proposals have been released by China, highlighting the need to balance innovation with safety. Meanwhile, Chinese research institutions, scientists, and policy experts have taken an active part in global discussions on AI safety.
然而在当前的地缘政治环境下，这种合作意愿常被外界所忽略。当前国际社会对AI的关注多集中于芯片、算力和基础设施竞争，但在这类热点话题之外，中美专家实际上仍在持续认真地就AI风险进行交流。尽管中美关系紧张，但过去几年中，两国科学家、研究人员和政策专家通过二轨对话机制保持着低调的沟通。这些对话涉及从“失控风险”、AI驱动的网络攻击、AI与生物安全风险，到术语定义和信任建设措施等多个议题。
*二轨对话（Track II Dialogue / Track II Diplomacy，1982年由美国外交官蒙特维尔提出）
一轨=官方政府谈判；
二轨=非官方智库、学者、退休官员、民间专家之间的非正式对话。
一、一轨vs二轨核心区别
1.一轨对话（Track I）
各国现任外交官、政府官员正式会谈，发言代表本国官方立场、受外交纪律约束，措辞严谨，敏感话题没法放开试探。
2.二轨对话（Track II）
参会人：智库研究员、高校学者、退役外交官、行业专家、NGO人士，不带现任公职身份，不代表本国政府正式表态 。
二、二轨的核心用处
1.探路试水：官方不方便开口的敏感议题（地地缘、军控、经贸分歧），二轨可以放开聊、试探底线，不用承担外交追责 ；
2.缓冲破冰：两国官方关系僵持、谈不拢时，靠二轨先拉近距离、积累共识；
3.成果落地：二轨聊出可行方案后，提交给本国官方，转化成一轨正式政策与协议。
三、补充小知识
还有Track 1.5（一轨半）：在职官员私下以非公务身份参会，现在国际交流最常用。
This point is often neglected amid today's geopolitical tensions. Much of the current international AI discourse focuses on strategic competition in chips, computing power, and infrastructure. However, beyond these headline issues, there is a less visible reality: Chinese and American experts are still engaging in serious discussions on AI risks. Despite the Sino-US tensions, Track II dialogues involving scientists, researchers, and policy experts from both countries have continued in a low-key manner over recent years. These dialogues cover a wide range of topics, from loss-of-control risks and AI-enabled cyber threats to AI and biosecurity risks, as well as terminology alignment and confidence-building measures.
在参与这些交流的过程中，我观察到，尽管双方在政治立场上存在差异，但中美专家对AI未来发展路径的担忧却高度相似。越来越多的人认识到，一些风险的规模已超出任何单一国家的应对能力。
I have personally observed that, despite political differences, experts from China and the US share remarkably similar concerns about the future trajectory of AI development. There is a growing awareness that some risks are simply too large for any single country to manage alone.
中美两国都有责任共同应对AI相关的风险。作为全球两个领先的AI国家，北京和华盛顿的决策不仅将影响技术的未来，也将深刻塑造全球AI风险的格局。
Both China and the United States share the responsibility to address AI-related risks. As the world's two leading AI powers, the decisions made in Beijing and Washington will not only influence the future of technological development but also significantly shape the global AI risk landscape.
由图灵奖得主Yoshua Bengio牵头、100多位AI专家共同撰写的最新《国际人工智能安全报告》指出了两个尤其值得关注的风险领域。
The latest International AI Safety Report, led by Turing Award winner Yoshua Bengio and co-authored by over 100 AI experts, highlights two particularly concerning risk areas.
*disconcerting(adj.)令人不安的、令人困惑的