Intelligent agents are intelligent systems capable of autonomous perception, memory, decision-making, interaction and execution, and represent a key form of artificial intelligence products and services. With the rapid development of new-generation artificial intelligence technologies such as large language models, intelligent agents are accelerating their deep integration with cyberspace and the physical world, profoundly transforming human lifestyles, modes of production and models of social governance. These Implementation Guidelines are formulated to implement the State Council's "Opinions on Deepening the Implementation of the 'AI+' Initiative" and to promote the standardised application and innovative development of intelligent agents.
I. Basic Principles
Guided by the objectives of advancing scientific and technological innovation, enhancing governance capabilities, building an industrial ecosystem, and improving people's well-being, we shall adhere to the principle of safety and controllability. The safety, reliability and trustworthiness of intelligent agents shall be established as fundamental requirements for development, running through the entire process of technological research and development, application deployment and promotion, to effectively prevent systemic risks.
We shall adhere to standardised and orderly development, adapting to the laws of technological evolution in intelligent agents. We shall establish a governance system that seamlessly integrates with existing policies and regulations, promotes industry self-regulation and autonomy, and clearly defines bottom lines and red lines, thereby advancing the practical application of intelligent agents in an orderly manner.
We shall uphold innovation-driven development, strengthening the synergy between theoretical, technological and engineering innovation. We shall achieve systematic breakthroughs in key intelligent agent technologies, improve the collaborative mechanism among government, industry, academia, research and end-users, build an open and shared ecosystem for intelligent agents, and enhance the vitality of industrial innovation.
We will adhere to an application-led approach, focusing on practical needs such as scientific research, industrial development, boosting consumption, public welfare and social governance. We will leverage the demonstration effect of typical application scenarios, proceeding from the simple to the complex in a step-by-step manner, to facilitate the technical validation, product iteration and practical implementation of intelligent agents.
II. Consolidating the Foundations for Development
We will consolidate the technological foundation, improve the standards system, and lower the barriers to R&D, adaptation and application of intelligent agents, thereby laying the groundwork for a diverse range of intelligent agent products and services.
(I) Improving the Technological Foundation
- Strengthen R&D in foundational technologies. Continuously enhance the performance of general-purpose foundational models, support the development of specialised models for specific industry sectors, and establish a model product matrix suitable for diverse scenarios and devices. Enhance the supply of high-quality datasets for the training and operation of intelligent agents. Intensify research and development efforts in areas such as task understanding, task planning, tool usage, long-term memory, mutual recognition and interoperability, and group collaboration to improve the generalisation capabilities of intelligent agents.
- Refine the agent toolchain. Conduct research into underlying agent frameworks, accelerate the development of key components such as perception, memory, decision-making, interaction and execution, and refine the toolchain for agent R&D, testing, deployment and operations. Develop security and governance tools such as adversarial sample detection and behavioural anomaly detection to enhance the ability to detect, intervene in, block and recover from non-compliant agent behaviour.
(II) Establishing Standard Protocols
- Establish a standardisation system for intelligent agents. Formulate guidance documents for standardisation work on intelligent agents and develop a standard framework. Systematically establish standardisation systems covering key technologies, major products, data exchange, application scenarios, quality assessment, security assurance and trusted authentication. Accelerate the formulation of foundational standards for interfaces between intelligent agents and software tools, application services and hardware peripherals. Strengthen the promotion and application of key national and industry standards for intelligent agent interconnection, such as the Intelligent Agent Interconnection Protocol (AIP). Support the formulation of mandatory standards in sectors such as healthcare, transport, media and public safety. Encourage enterprises to develop products and services in accordance with relevant standards to enhance the standardisation of intelligent agents. Actively participate in the formulation of international standards.
- Plan and develop the intelligent internet. Research the establishment of an intelligent internet architecture and explore the creation of an intelligent agent registration platform to provide services such as digital identity management, search and discovery, and capability declaration for intelligent agents, whilst supporting the querying and management of information regarding developers, deployment methods, interface protocols and compliance certification. Enhance multi-agent collaboration capabilities and research foundational technologies such as agent identity identification, trusted interconnection, compliant payment, security protection and conflict resolution. Leverage the technical advantages of Internet Protocol version 6 (IPv6) to enhance end-to-end communication capabilities of intelligent agents. Explore the establishment of a monitoring indicator system for the intelligent internet.
III. Upholding the Security Baseline
Adhere to the principles of putting people first, intelligent systems for the greater good, multi-stakeholder governance, and safety and stability. Foster an institutional environment that promotes standardised development and encourages innovation, thereby facilitating the healthy and orderly development of intelligent agents.
(I) Clarifying Product Guidelines
- Refine policies, regulations and ethical standards. Accelerate research into policies, regulations and ethical standards pertaining to intelligent entities. Leverage the content resources and vetting capabilities of professional institutions to ensure that the behaviour of intelligent entities complies with laws, regulations and mainstream values. Prevent intelligent entities from exploiting their data advantages or personification technologies to disseminate harmful values or engage in algorithmic exploitation. Mitigate risks such as addiction and emotional dependency among minors and the elderly. Ensure seamless integration with systems such as AI ethics review mechanisms.
- Clarify decision-making authority. Subject to compliance with laws and regulations and respect for public morality and ethical standards, delineate the reasonable boundaries and required authorisations for various decision-making methods, including decisions reserved solely for the user, decisions requiring user authorisation, and autonomous decisions by AI agents. Ensure that users retain the right to be informed and the final decision-making authority regarding autonomous decisions made by AI agents; AI agents must not execute operations beyond the scope of user authorisation.
- Strengthen behavioural control. Develop technologies such as rule embedding and behavioural fencing to ensure that the behaviour of intelligent agents in public spaces, private spaces and designated areas is lawful and compliant. Explore the use of technologies such as blockchain to establish mechanisms for verifying and tracing the behaviour of intelligent agents in critical application scenarios, thereby preventing major risks arising from inappropriate behaviour.
(II) Mitigating Security Risks
- Enhance intrinsic security capabilities. Research security technologies covering data security, personal information protection, cryptographic protection, attack detection, access control and behavioural control for intelligent agents, to enhance the security safeguards of intelligent agent systems and mitigate security risks such as data poisoning, privacy breaches, algorithm tampering, system vulnerabilities and operational runaway. Research security detection technologies for intelligent agents and explore the establishment of a security assessment framework for them.
- Strengthen supply chain security. Formulate full-lifecycle security standards covering the development, deployment, application and maintenance of intelligent agents, and strengthen security management in areas such as model integration, API calls and the use of extension tools. Explore the establishment of a security information sharing and early warning mechanism for the intelligent agent supply chain, issue risk alerts in a timely manner, and enhance security safeguards.
- Mitigate risks arising from applications. Improve routine risk identification, early warning and intervention mechanisms for intelligent agents, and strengthen risk management capabilities such as human-machine collaborative review and interception to prevent systemic security risks. Strengthen security management of intelligent agent applications to prevent their use in illegal activities such as automated attacks, privacy infringements, the generation and dissemination of false information, and online fraud.
(III) Improving the Governance System
- Establish a categorised and tiered governance framework. Based on application scenarios and potential impacts, implement a prudent and steady tiered governance approach for intelligent agents. For sensitive sectors and key industries, the Cyberspace Administration, in conjunction with the relevant industry authorities, shall determine permissible application scenarios and implement management measures—such as filing, testing, and the recall of problematic products—in accordance with relevant laws, regulations, regulatory requirements, and security standards. For certain low-risk areas, such as lifestyle and entertainment or daily office work, improve assessment and testing tools for intelligent agents to achieve efficient governance through compliance self-assessment, information reporting, distribution platform management, and industry self-regulation.
- Improve the compliance service system. Strengthen the provision of professional services such as risk monitoring and early warning, testing and evaluation, consultancy and certification for intelligent agents, and guide the industry to actively develop monitoring tools for intelligent agents. Conduct third-party evaluation services covering the functionality, performance, quality and compliance of intelligent agents, and promote the mutual recognition of certification and testing results to provide guidance for users selecting intelligent agents. Compile and publish reports on the technical and application maturity of intelligent agents to provide guidance for industrial research, development and application.
(IV) Strengthening Industry Self-Regulation
- Guide the industry to strengthen self-regulation. Encourage industry organisations and major enterprises to jointly formulate industry self-regulatory rules, clarifying detailed regulations on AI functionality compliance, algorithm governance, intellectual property protection and fair competition. Guide AI development platforms, distribution platforms and service providers to establish fair and reasonable platform rules, user service agreements and privacy policies, clarifying the rights and responsibilities of both supply and demand sides to ensure the healthy development of the industry. Strengthen public education on the risks of AI applications to enhance users' security awareness.
- Explore credit rating mechanisms. Guide industry organisations to establish voluntary credit rating mechanisms for market entities in the intelligent agent sector, conducting credit assessments for behaviours such as misuse of technology, inducing consumption, false advertising and concealing information on defects, and implementing sanctions for dishonest conduct in accordance with laws and regulations. Encourage intelligent agent developers, development platforms, distribution platforms and service providers to participate in credit ratings, jointly fostering a favourable development environment.
IV. Strengthening Application-Driven Development
Actively and prudently promote the application of intelligent agents in typical scenarios to drive the optimisation and enhancement of technical products, and explore the formation of replicable and scalable models for the practical implementation of intelligent agents.
(I) Scientific Research
- Research and exploration. Develop intelligent agents for theoretical deduction and simulation to identify potential technical pathways. Strengthen capabilities in information integration and knowledge system construction to enhance research capabilities in the natural sciences, philosophy and the social sciences. Promote the integration of intelligent agents with scientific instruments and experimental platforms to achieve full-process intelligence in areas such as scheme design, experimental operations, data processing and result analysis.
- R&D support. Develop software development intelligent agents to enhance full-process development capabilities, including requirements analysis, architectural design, code generation and testing. Promote the integration of intelligent agents with software such as Computer-Aided Design (CAD) and Computer-Aided Engineering (CAE) to provide functions such as design scheme generation, simulation verification and parameter optimisation.
(II) Industrial Development
- Intelligent manufacturing. Research and develop production management agents to dynamically optimise production scheduling, resource allocation and process sequencing; promote the application of agents in the industrial internet sector to enhance the level of refined management within enterprises. Improve the agents' capabilities in optimising process parameters, verifying machining accuracy and identifying product defects; promote the integration of agents with CNC machine tools, industrial robots and automated production lines to enhance quality, improve efficiency and reduce costs.
- Energy and resources. Research and develop intelligent agents for sensing environmental factors such as air, water, soil and noise, to enhance early warning capabilities for risks such as natural disasters and environmental pollution. Rely on intelligent agents to strengthen full-cycle management capabilities for national land and spatial resources. Utilise intelligent agents to achieve efficient exploration of energy, metal and other mineral resources. Develop intelligent agents for power dispatch, electricity consumption monitoring and power grid maintenance to improve the efficiency of power resource utilisation.
- Transport. Develop intelligent agents for traffic safety supervision and emergency command and dispatch to improve capabilities in identifying traffic violations, issuing early warnings for risks to transport infrastructure, monitoring key vehicles (and vessels), and responding rapidly to accidents. Optimise the performance of intelligent agents for traffic monitoring and dispatch, and develop intelligent agents for the control of transport vehicles to enhance traffic efficiency across road, waterway and airspace networks.
- Agricultural production. Develop intelligent agents for agricultural services to provide technical guidance, as well as pest and disease diagnosis and control. Promote the application of intelligent agents in crop cultivation, livestock farming and high-efficiency breeding to advance the intelligent transformation of agriculture. Facilitate the integration of intelligent agents with smart agricultural machinery, smart greenhouses and agricultural service platforms to enhance agricultural production efficiency.
- Financial services. Develop intelligent agents for financial risk control to enhance risk identification capabilities in areas such as credit approval, transaction monitoring and account security. Refine the agents' anomaly detection and compliance audit functions to improve capabilities in credit default prediction, credit card fraud interception and anti-money laundering monitoring.
(III) Boosting Consumption
- End-user applications. Promote the use of intelligent agents to empower internet applications and services, optimising the user experience for online shopping, travel navigation, utility bill payments and daily office tasks. Promote the coordinated development of AI agents with terminal devices such as mobile phones, computers, cars, home appliances, wearables and consumer-grade robots, to enhance the ability to complete tasks across applications and devices.
- Culture and tourism. Develop AI agents for content creation in literature, music, painting, audiovisual media and performing arts to promote the dissemination of outstanding cultural works. Develop AI agents for tourism services such as smart guided tours, multilingual translation, and services suitable for the elderly and disabled to improve the standard of tourism services.
- Commercial services. Enhance the customer service capabilities of intelligent agents to provide 24/7 consultation, booking and after-sales services. Develop embodied intelligent agents for guidance, cleaning, warehousing and distribution to improve operational efficiency in commercial venues such as catering, retail, accommodation and logistics. Explore the provision of low-cost domestic help, elderly care, childcare and disability support services through embodied intelligent agents.
(IV) Public Welfare
- Education and teaching. Explore the development of intelligent agents for lesson material generation, homework marking and learning progress analysis to improve teachers' work efficiency. Utilise intelligent agents to develop personalised learning plans and refine functions such as intelligent study guidance, Q&A support and virtual teaching assistants. Support online education platforms in developing intelligent agents to provide lifelong learning services.
- Healthcare. Enhance the performance of medical assistance agents in medical imaging analysis, disease diagnosis reasoning and the generation of personalised treatment plans. Explore agents for medication management, surgical scheduling and medical record management to improve the efficiency of healthcare services. Steadily develop agents for pre-consultation and report analysis to enhance the patient experience.
- Human resources. Explore the application of agents in employment promotion, the training and assessment of technical and skilled personnel, and public services for labour relations to enhance employment service capabilities. Develop AI agents for social insurance, labour dispute arbitration and the management of wage arrears to safeguard the legitimate rights and interests of workers.
- Information services. Explore the application of AI agents in the construction and management of online content. Encourage information-releasing departments and content distribution platforms to develop AI agents for user analysis, topic planning, editorial processing, distribution and recommendation, intelligent review, public opinion guidance, emotional support and real-time translation, thereby achieving the efficient integration of multimodal and cross-domain information.
(V) Social Governance
- Public administration services. Explore the use of AI agents to assist with administrative approvals, promoting the intelligent transformation of government approval processes. Develop policy consultation AI agents to provide round-the-clock online government advisory services and procedural guidance. Explore the proactive delivery of tailored policy information, service reminders and application guides, accelerating the transition from a model where "people seek services" to one where "services find people".
- Judicial services. Explore AI agents for end-to-end case handling assistance to enhance capabilities in organising case materials, entering case information, reviewing evidence, and assisting with the generation of legal documents. Develop AI agents for legal publicity, legal consultation and legal supervision to provide the public with efficient and convenient online judicial services.
- Public safety. Explore AI agents for monitoring and early warning, emergency response, rescue coordination and collaborative governance to enhance capabilities in work safety supervision, disaster prevention, mitigation and relief. Enhance the capabilities of intelligent agents in identifying abnormal behaviour, issuing early warnings for potential threats, and implementing dynamic prevention and control measures to safeguard public security. Promote the practical application of embodied intelligent agents in areas such as disaster relief, security patrols and inspections, and the handling of hazardous materials.
- Urban governance. Explore the application of intelligent agents in urban planning, construction and governance to support smart construction, building management and the safe operation of urban infrastructure, thereby raising the professional standards of urban governance and improving the quality of the urban living environment.
- Tendering and bidding. Explore the use of intelligent agents in tendering and bidding to achieve end-to-end intelligent management of these activities, ensuring standardised and efficient processes throughout. Enhance the level of intelligence in tendering and bidding transactions, services and supervision, to ensure scientific and reasonable tendering, fair and impartial bid evaluation, and efficient, transparent supervision.
V. Building an Innovation Ecosystem
Facilitate the flow of supply and demand, promote high-level interaction between the R&D and demand sides, and foster an intelligent agent industry ecosystem driven by market forces and internal momentum.
(I) Promoting Industrial Cooperation
- Foster open-source innovation. Guide domestic artificial intelligence open-source communities to strengthen their deployment of intelligent agents and carry out compatibility and adaptation work between intelligent agents and open-source chips, open-source operating systems, and open-source large models. Guide enterprises, universities and research institutions to actively participate in open-source projects such as intelligent agent frameworks, interaction interfaces and toolchains, promote the integrated development of technical systems, and accelerate the enhancement of international influence.
- Establish industry collaboration platforms. Leverage the role of industry collaboration platforms such as intelligent-agent ecosystem alliances and technology verification laboratories to coordinate upstream and downstream supply-chain participants in joint R&D on common intelligent-agent technologies, standards-setting, and assessment and certification work. Conduct training of composite talent combining intelligent-agent technology with industrial application. Guide enterprises in fields such as internet applications and smart terminals to jointly build the ecosystem and explore mutually beneficial cooperation models.
(II) Strengthening Application Promotion
- Build application promotion channels. Drive the establishment of intelligent-agent software stores and industry supply-demand information platforms, guiding intelligent-agent enterprises to actively release products and form clustering effects. Conduct intelligent-agent application supply-demand matching activities, attracting intelligent-agent enterprises to undertake customised product development through mechanisms such as open tendering and the "unveil-and-take-the-helm" (揭榜挂帅) challenge model. Guide hardware-system and software enterprises to develop products and services built on intelligent agents, cultivating the user market.
- Advance the opening of key scenarios. Drive the opening of intelligent-agent application scenarios in key sectors; conduct intelligent-agent application pilots in industrial clusters, key industries and key sectors; build a portfolio of demonstration projects with a leading role. Develop market-based, professional intelligent-agent technology-transfer service organisations; explore intelligent-agent application scenarios; improve the efficiency of technology-results conversion. Promote industry data sharing and openness to support the training and deployment of intelligent agents in key scenarios.
- Actively cultivate the global ecosystem. Leveraging international platforms such as the World Artificial Intelligence Conference and the World Internet Conference, exchange and showcase intelligent-agent technology innovation achievements. Promote the adaptation of intelligent agents by terminal-device and software enterprises; guide relevant enterprises to undertake overseas compliance work; promote the adaptation of intelligent agents to local laws, regulations and cultural customs.
VI. Safeguarding Measures
The Cyberspace Administration of China, the National Development and Reform Commission, and the Ministry of Industry and Information Technology, in conjunction with other relevant parties, shall strengthen overall planning, intensify resource integration and the coordination of capabilities, refine supporting policies, build a coordinated working force, and drive the implementation of key tasks. We shall establish and improve an evaluation indicator system for the development of intelligent agents, and strengthen the monitoring, evaluation, rolling implementation and dynamic adjustment of the standardised application and innovative development of intelligent agents.