Company News

Our reporter Cui Shuang

 

More and more daily tasks such as taking a taxi, ordering meals, making travel plans, and organizing meeting minutes are being "delegated" by intelligent agents. How to make them identify each other and work together safely has become a new topic that must be answered in the development of intelligent Internet.

 

On November 4, the 2026 World Internet Conference Wuzhen Summit "Global cooperation and innovation of intelligent agents for mutual trust, interconnection, interoperability - building an intelligent Internet forum together" was held. The Forum released the Smart Internet Innovation and Development Cooperation Initiative (hereinafter referred to as the Initiative) to promote all parties to explore the development path of smart Internet.

 

"As a key force to promote the development of intelligent Internet, agents are accelerating the evolution from dialogue and interaction to independent implementation, from local assistance to full process collaboration, and from single operation to cross domain mutual trust, interconnection, and interoperability." Wang Jingtao, deputy director of the National Internet Information Office, said at the forum that he hopes to further build consensus among all parties to build an intelligent Internet that is native, open, innovative, secure, credible, and inclusive.

 

The network technology architecture urgently needs to be upgraded

 

The development of intelligent agents first brings about changes in the objects and scale of network connections.

 

"In the past, the main objects of Internet connection were people, machines and things. In the future, it will also include agents, as well as data, models, skills and other elements. ”Wang Xiaoyun, an academician of the CAE Member, chief engineer and chief scientist of China Mobile, said that the connection scale could expand from hundreds of millions to hundreds of billions or even trillions.

 

The increase in connected objects and the expansion of collaboration scale pose new requirements for network capabilities. Wang Xiaoyun believes that the paradigm of network services will shift from the past "best effort" to providing differentiated guarantees for high-value businesses, with higher requirements for indicators such as latency, jitter, and reliability. In the process of model training and reasoning, intelligent computing centers need to be interconnected, and network protocols need to adapt to new business requirements.

 

"This year can be said to be the first year of agents. The emergence of agents has put forward new demands for Internet technology architecture." Wang Jingtao said that we should adhere to innovation driven, increase investment in research and development of key common technologies such as new network communication theory and intelligent native network architecture, break through technologies such as agent environment perception, complex reasoning, long-term planning, and reliable execution, and improve the level of network intelligence.

 

To this end, the "Initiative" proposes to deepen technological innovation and drive network evolution; Building infrastructure together to facilitate ubiquitous connectivity. We will work together to promote the collaborative deployment and interconnection of new generation information infrastructures such as IPv6, 5G-A/6G, satellite Internet, and intelligent computing, and provide support for the popularization of intelligent applications and services.

 

Accelerate the establishment of mutual recognition and trust mechanism

 

If network infrastructure solves how intelligent agents connect, then mutual trust mechanisms need to answer how intelligent agents can collaborate securely. "Intelligent Internet is not the superposition application of artificial intelligence technology on the network, but the systematic reconstruction of network architecture, resource organization and service form," said Bo Zhaoyi, deputy director and senior engineer of the Data and Technology Support Center of the Central Cyberspace Office.

 

He proposed that a large number of intelligent agents require identification, addressing, and mutual discovery, and networks not only need to solve the problem of "connectivity", but also the problem of "recognition and trustworthiness". Therefore, identity identification and interoperability protocols have become the basic technical issues for the development of intelligent Internet.

 

Chen Liang, vice president of Ant Group, chief technology officer of Alipay and the big security business group, also believes that the era of agent needs to build a trust infrastructure to allow identity, authorization and responsibility to enter every connection.

 

The upper limit of the scale of intelligent agents is not 'whether they can complete a task', but 'whether they can be entrusted with a task', "Chen Liang said. Only by establishing a trustworthy identity and authorization mechanism, clarifying the behavior boundaries and responsibility relationships of intelligent agents, can conditions be created for intelligent agents to undertake more complex tasks and carry out more extensive cooperation.

 

The autonomy and dynamic changes of intelligent agents also bring new challenges to security governance. Wu Hequan, an academician of the CAE Member, pointed out that compared with applications, intelligence can learn by itself and keep changing, and it may also penetrate and call other agents, bringing difficulties to supervision and responsibility identification.

 

The Initiative proposes to jointly develop open, non discriminatory, and transparent international standards for intelligent agent interoperability, and establish a mutual trust framework covering the entire lifecycle of intelligent agent discovery, negotiation, and execution. By establishing standards and rules, intelligent agents developed by different platforms and entities can collaborate on a trustworthy basis, reducing ecological fragmentation and redundant investment.

 

On the forum, the intelligent agent identity registration and resolution platform was officially released, focusing on solving identity encoding, cross domain resolution, and trusted interoperability issues, supporting interconnection and collaboration between different intelligent agents. The forum also released the "Self discipline Convention for the Development of Intelligent Agents", proposing eight self-discipline measures, covering the improvement of intelligent agent trust system, the construction of intelligent agent interconnection infrastructure, and other aspects. (Jiaxing Science and Technology Daily, November 4th)

 

(