人民网北京10月23日电，22日，在人民网主办的“认知·具身·未来——具身认知与AI进化交流会暨成果发布会”上，具身智能应用及安全伦理联盟成立，并召开首次专题座谈会。该联盟由人民网依托传播内容认知全国重点实验室，联合高校、科研院所和产业链企业共同组建，作为主流价值语料生态联盟的延伸与拓展，纳入其生态建设体系。
People's Daily Online, Beijing, October 23 — On the 22nd, at the "Cognition · Embodiment · Future — Embodied Cognition and AI Evolution Exchange Conference and Achievement Release" hosted by People's Daily Online, the Embodied Intelligence Application and Safety Ethics Alliance was established and held its first special symposium. The alliance was jointly formed by People's Daily Online relying on the National Key Laboratory of Communication Content Cognition, together with universities, research institutes, and supply chain enterprises, as an extension and expansion of the Mainstream Value Corpus Ecosystem Alliance, incorporated into its ecosystem construction system.
据介绍，联盟主要聚焦应用的安全伦理，重点打造四个平台。安全治理的实践平台，开展安全伦理框架研究，推动物理技术安全、数据隐私保护、行为规范、责任认定等行业标准和国家标准制定，建设安全评测的验证体系。标准规则制定的引领平台，推进语料采集、标注、交换、应用全链条标准体系建设。产业协同的枢纽平台，建立常态化政产学研用沟通机制，让需求、供给、政策及时互通。人才培养的支撑平台，联合培养具身智能复合型人才，为产业发展持续输送高质量人才。
According to the introduction, the alliance mainly focuses on application safety ethics, building four key platforms. The safety governance practice platform conducts safety ethics framework research, promotes industry and national standards including physical technology safety, data privacy protection, behavioral norms, and responsibility determination, and builds a safety evaluation validation system. The standard-setting leading platform advances the full-chain standard system for corpus collection, annotation, exchange, and application. The industrial collaboration hub platform establishes normalized government-industry-academia-research-application communication mechanisms for timely exchange of demand, supply, and policy. The talent cultivation support platform jointly cultivates embodied intelligence interdisciplinary talent, continuously supplying high-quality talent to industry development.
在联盟伙伴方面，学术端重点邀请人工智能、具身智能、数据治理等领域的优势高校和科研院所，为联盟提供坚实的学术支撑；产业端广泛吸纳数据、模型、应用等产业链上下游企业机构，推动技术与场景、供给与需求精准对接。活动现场，18家联盟成员单位代表共同上台启动，具身智能应用及安全伦理联盟正式成立。联盟成员单位包括传播内容认知全国重点实验室、自主智能无人系统全国重点实验室、人民网人工智能研究院、工信部电子五所人工智能中心、北京市人工智能协会、北控智科数智研究院、朝阳科技集团、地平线公司、地瓜机器人、光轮智能、京东、节卡机器人、灵心巧手、乐聚机器人、面壁智能、尚品宅配、英利能源、中关村机器人产业创新发展有限公司。
On alliance partners, the academic side focuses on inviting leading universities and research institutes in AI, embodied intelligence, and data governance to provide solid academic support; the industrial side broadly absorbs upstream and downstream enterprises and institutions in data, models, and applications, promoting precise matching between technology and scenarios, supply and demand. At the event, representatives from 18 alliance member units jointly launched on stage, officially establishing the Embodied Intelligence Application and Safety Ethics Alliance. Member units include the National Key Laboratory of Communication Content Cognition, the National Key Laboratory of Autonomous Intelligent Unmanned Systems, People's Daily Online AI Research Institute, MIIT Electronic Fifth Institute AI Center, Beijing AI Association, Beikong Zhike Digital Intelligence Research Institute, Chaoyang Technology Group, Horizon, D-Robotics, Guanglun Intelligent, JD.com, JAKA Robotics, Linker Hand, Leju Robotics, ModelBest, Shangpin Home Delivery, Yingli Energy, and Zhongguancun Robot Industry Innovation Development Co., Ltd.
这个联盟的成立，和前面讨论的机器人伤人责任划分、德国警方逮捕涉嫌性骚扰的机器人、芝加哥外卖机器人侵犯生物识别隐私、董重青说机器人进家庭最大障碍在法律法规伦理层面，是同一类问题在产业层面的回应。前面讨论过武汉机器狗撞伤两岁女童的责任划分问题，也讨论过Character.AI被肯塔基州起诉聊天机器人怂恿用户自残，还讨论过欧盟对AI伦理的监管框架。这些案例都指向同一个现实，具身智能从实验室走向真实场景，技术问题只是一部分，安全、隐私、责任、伦理这些配套体系必须同步跟上。联盟的四个平台里，安全治理的实践平台和标准规则制定的引领平台，直接对应这个需求。
The alliance's establishment responds at the industry level to the same class of problems as the previously discussed robot injury responsibility division, German police arresting a robot suspected of sexual harassment, Chicago delivery robots violating biometric privacy, and Dong Zhongqing saying the biggest obstacle to robots entering homes is at the legal, regulatory, and ethical level. Earlier we discussed the responsibility division question in Wuhan where a robot dog injured a two-year-old girl, Character.AI being sued by Kentucky over chatbots encouraging self-harm, and the EU's AI ethics regulatory framework. These cases all point to the same reality—as embodied intelligence moves from laboratory to real scenarios, technical problems are only part of it; safety, privacy, responsibility, and ethics supporting systems must keep pace simultaneously. Among the alliance's four platforms, the safety governance practice platform and standard-setting leading platform directly address this need.
联盟成员单位的构成也值得看一眼。18家单位里，有传播内容认知全国重点实验室、自主智能无人系统全国重点实验室这样的国家级科研平台，有人民网人工智能研究院、工信部电子五所人工智能中心这样的官方研究机构，有地平线、地瓜机器人、光轮智能、面壁智能这样的技术公司，有京东、节卡机器人、灵心巧手、乐聚机器人这样的产业方，还有尚品宅配、英利能源这样的应用方。这个组合的特点是，把技术方、产业方、应用方、研究机构和媒体平台都拉到了一张桌子上，而不是只让技术公司自己讨论伦理。前面讨论过爱思科技的Agnes Harness框架、亮源新创的泛化能力、OpenLoong的开源体系，这些都是在技术层面找解法。联盟做的是另一件事，在规则层面找共识。技术跑得快，规则跟不上，行业就会在真实场景里反复踩坑。联盟能不能把安全伦理从口号变成可执行的标准和验证体系，取决于它接下来能不能拿出具体的标准文本和评测方法。
The composition of alliance member units also deserves a look. Among the 18 units are national-level research platforms including the National Key Laboratory of Communication Content Cognition and the National Key Laboratory of Autonomous Intelligent Unmanned Systems, official research institutions including People's Daily Online AI Research Institute and MIIT Electronic Fifth Institute AI Center, technology companies including Horizon, D-Robotics, Guanglun Intelligent, and ModelBest, industrial parties including JD.com, JAKA Robotics, Linker Hand, and Leju Robotics, and application parties including Shangpin Home Delivery and Yingli Energy. This combination's characteristic is bringing technology parties, industrial parties, application parties, research institutions, and media platforms to one table rather than letting technology companies discuss ethics alone. Earlier we discussed Agnes AI's Agnes Harness framework, Liangyuan Xinsheng's generalization capability, and OpenLoong's open-source system—all seeking solutions at the technical level. The alliance does another thing—seeking consensus at the rule level. Technology runs fast; if rules cannot keep up, the industry will repeatedly step into pits in real scenarios. Whether the alliance can turn safety ethics from slogans into executable standards and validation systems depends on whether it can produce specific standard texts and evaluation methods next.
