A specialized forum focusing on the security of embodied artificial intelligence was successfully convened in Beijing on the afternoon of June 2nd, as part of the broader Beijing Cyber Security Conference (BCS 2026).
Held at the China National Convention Center under the theme "Embodied Intelligence in Action, Security as the Foundation," the event brought together government officials, industry specialists, and representatives from leading enterprises.
Participants engaged in in-depth discussions on the security challenges and opportunities arising as embodied intelligence transitions from laboratory research to widespread adoption across numerous industries.
A highlight of the forum was the signing of strategic cooperation agreements between QiAnXin Group and six prominent industry leaders.
These agreements aim to establish baseline security standards for the sector and collaboratively build a secure and beneficial industrial ecosystem, thereby constructing a robust security defense for embodied intelligence, a key driver of new quality productive forces.
In his opening remarks, Lu Wei, Vice Chairman of the Cyberspace Association of China, emphasized that embodied intelligence is propelling AI from virtual spaces into the physical world, with its risk profile evolving from traditional content-related hazards to autonomous action risks.
He stressed that without a solid security foundation, the large-scale deployment of embodied intelligence would remain an unattainable ideal.
Lu Wei advocated for adhering to a full lifecycle security philosophy, establishing a zero-trust security protection architecture, and shifting governance from post-incident response to a closed-loop, lifecycle approach.
Qi Xiangdong, Chairman of QiAnXin Group, stated that 2026 marks the "inaugural year" for embodied intelligence, with technological readiness and market conditions maturing, yet the associated security issues are particularly severe.
He noted that embodied intelligence serves as a critical bridge linking cyberspace and the physical world, meaning that successful attacks could transform data breaches into tangible physical disasters.
Qi Xiangdong underscored the necessity of leveraging technological innovation to address the novel risks emerging from this virtual-physical fusion, expressing QiAnXin's commitment to collaborating with the industry to contribute Chinese solutions to global intelligent agent security governance.
The forum featured a formal signing ceremony for strategic cooperation in embodied AI security.
Chen Huaping, Vice President of QiAnXin Group, represented the company in forging partnerships with representatives from six leading embodied AI enterprises: Geek+, ZhiShen Technology, Beijing Humanoid Robot Innovation Center, Yunji Technology, Guangzhou Muli Technology, and Zhongguancun Robot Industry Innovation and Development Co., Ltd.
This collaboration is designed to define industry security baselines and safeguard the entire trusted chain of embodied intelligence from perception to action, signifying a shift from isolated point defenses to comprehensive, system-wide trusted protection and opening a new chapter of synergistic development for the industry.
The keynote speech segment saw multiple industry experts take the stage, delivering a feast of ideas and technology as they dissected, from various angles including technical implementation, data-driven strategies, and offensive/defensive practices, the security hurdles that must be overcome as embodied intelligence enters its anticipated "golden decade."
Li Hongbo, CTO of Geek+, presented on the "Path from Innovation to Trustworthiness: Securing the Deployment of Embodied AI Robots."
Drawing on Geek+'s practical experience in warehousing and logistics, he explained how AI robots are reshaping production paradigms through "embodied cerebellum" and "embodied cerebrum" architectures.
Li Hongbo identified the core security challenges for embodied intelligence as residing in two dimensions: Safety (physical safety) and Security (information security).
He proposed an endogenous security philosophy of "using magic to defeat magic," which involves building inherent security into the technology itself through robust hardware and clean-slate algorithms, while simultaneously establishing a comprehensive, in-depth active defense system to ensure robots operate both efficiently and reliably.
Liu Yulong, Co-founder of ZhiShen Technology, delivered a speech titled "Empowering a Myriad of Industries: How to Jointly Build the Security Foundation for Embodied Intelligence."
He detailed ZhiShen's technological expertise in quadruped robots and embodied intelligence "cerebellum/cerebrum" systems, along with its experience in delivering tens of thousands of units, while candidly addressing the early industry pain point of insufficient attention to information security from equipment manufacturers.
Liu Yulong argued that with product globalization and the implementation of regulations like the Cyber Resilience Act, companies must transform compliance red lines into design baselines.
He called on the industry to collaboratively build an "industrial-grade immune system," enhancing reliability through hardware encryption chips and establishing mechanisms for sharing vulnerability information to tackle multifaceted security challenges.
Xie Yunpeng, Vice President of Yunji Technology, centered his discussion on how security enables the demographic dividend for robots.
He posited that robots will replace simple, repetitive tasks rather than jobs in their entirety.
Xie Yunpeng emphasized that security is the cornerstone of an intelligent society, noting that Yunji Technology, in partnership with QiAnXin, has launched a high-security series of robots dedicated to addressing the vulnerabilities of embodied intelligence.
He urged the industry to focus on dual guarantees of physical stability and digital security, advocating for a management closed-loop of "goal-monitoring-traceability-feedback" to elevate robots from mere tools to stable and efficient laborers capable of operating in high-security-demand scenarios.
Wen Mingxing, Executive Director of the China-Singapore International Joint Research Institute, outlined implementation pathways for driving the large-scale adoption of embodied intelligence, focusing on the construction of a high-quality data flywheel.
He identified the scarcity of high-quality interactive data as the core bottleneck limiting industry scaling, noting that the data volume required for embodied intelligence far exceeds that for large language models and autonomous driving.
Industry data supply, he explained, comes from four main sources: real-world collection, simulation synthesis, manual control, and public data, each with its own cost and quality trade-offs, while simultaneously facing the dual obstacles of high real-world collection costs and the disconnect between simulated data and real scenarios.
His team's independently developed integrated platform uses a "REAL-SIM-REAL" technique to create an interactive data closed-loop, achieving a hundredfold increase in automated data generation efficiency.
Gong Yushan, CEO of QiAnXin Group's AI subsidiary, analyzed embodied intelligence security strategies from an offensive/defensive perspective, based on penetration testing results from a variety of robot types.
Following laboratory testing of mainstream humanoid, quadruped, and wheeled robots, the team identified eight major security risk dimensions: hardware supply chain, firmware systems, near-field communication, and cloud operations and maintenance, among others.
Leveraging these findings, QiAnXin has constructed a six-layer, in-depth defense system covering hardware, firmware, communication, and the cloud, while also launching a robot security assessment service for the entire industry to assist manufacturers in conducting full-chain security testing.
Concurrently, Gong Yushan called for the industry to unify security standards and clarify liability frameworks before large-scale deployment, using offensive/defensive exercises to continuously enhance the industry's resilience.
As the inaugural embodied AI security forum within the Beijing Cyber Security Conference, this event precisely targeted the critical window in 2026 as embodied intelligence begins its industrial-scale rollout, facilitating communication channels between security service providers, robot manufacturers, research institutions, and standards organizations.
All participants agreed that security must be the foremost priority in the development of embodied intelligence.
Through ecosystem co-creation, standard-setting, and technological innovation, the industry is joining hands to build a new, secure, controllable, and beneficial ecosystem for embodied intelligence, safeguarding the high-quality development of new quality productive forces.
Comments