In a significant step towards responsible AI development, 20 leading companies have officially unveiled the “Code for the Development of Intelligent Agents.” The document was presented on November 4th at the “Global Cooperation and Innovation for Intelligent Agent Interoperability – Building the Intelligent Internet Together” forum, part of the World Internet Summit in Wuzhen. Key signatories include the China Academy of Information and Communications Technology (CAICT), Huawei, Alibaba, Tencent, Xiaomi, Baidu, Volc Engine, Ant Group, and Douyin.
Core Principles of the Code
According to CAICT, this code represents a concrete initiative to implement the “strengthening of industry self-regulation” goals outlined in the “Guidelines for the Standardized Application and Innovative Development of Intelligent Agents.” The code establishes eight fundamental self-regulatory principles, encompassing:
- Enhancing the trust system for intelligent agents.
- Building infrastructure for the interconnectedness of intelligent agents.
- Improving interoperability levels between intelligent agents.
- Strengthening the security and protection of intelligent agents.
- Refining the management mechanisms for intelligent agents.
- Protecting the rights of intelligent agent users.
- Accelerating the adoption of intelligent agents across various industries.
- Promoting universal access to intelligent agent technologies.
Combating Malicious Use
The code places particular emphasis on management mechanisms. It advocates for the active development of technologies that enable the expression and integration of rules, as well as compliance testing against standards. The aim is to ensure that intelligent agent operations align with laws and societal norms. Crucially, the initiative seeks to collectively counter the use of intelligent agents for algorithmic coercion, fraud, and other illicit activities.
It’s important to understand that an Intelligent Agent (AI Agent) is an artificial intelligence system capable of perceiving its environment, making independent decisions, and executing tasks. Unlike conventional chatbots that respond only to single commands, AI agents can manage multi-step workflows with minimal human intervention.
Technological Accessibility and Collaborative Development
Regarding technological accessibility, the code encourages the collaborative development of open foundational and shared technologies, alongside participation in crafting international standards and regulations. Through joint research, personnel training, and other collaborative efforts, the goal is to create an innovative network benefiting developers worldwide. The objective is to continuously enhance ease of access and local deployment of these technologies, making intelligent agent services available to users in more countries and regions.
Future Outlook
Representatives from the organizations behind the code emphasized that this initiative is a crucial step towards the normalized development of the intelligent agent industry. The document is expected to foster consensus-building, strengthen industry self-regulation, and boost mutual trust, interconnectivity, and interoperability among intelligent agents. This, in turn, will help guide their standardized application and innovative growth.
The next phase will involve the founding companies committing to the code’s provisions and actively embracing oversight from industry organizations and the public. Meanwhile, all market players are invited to contribute to the development of the intelligent agent ecosystem and the code itself.
