China Proposes a Global AI Agent Trust and Interoperability Initiative
Identity, discovery, protocol interfaces, privacy, and security are becoming shared requirements for scaling AI agents.
Key takeaways
A global cooperation initiative released on July 17, 2026 calls for mutual trust, connectivity, and interoperability among AI agents. Its nine proposals span ecosystem development, security governance, open-source collaboration, standard interfaces, privacy protection, and digital inclusion. The initiative arrives alongside an ITU focus group on agent identity and access management, the Linux Foundation's Agent Name Service proposal, and the A2A protocol project. Together, these efforts show AI agent competition expanding beyond model quality toward trusted identity, service discovery, cross-platform communication, permission control, and auditable execution. For users and small teams, the immediate task is not to assume a universal standard exists, but to demand clear identities, scoped permissions, logs, approval gates, and rollback paths.
# China Proposes a Global AI Agent Trust and Interoperability Initiative
Published: July 22, 2026
Table of contents
- Direct answer
- Fact sources
- Definition, scenarios, steps, and risks
- Why it matters
- Impact for ordinary AI users
- Related tools/tutorials
- FAQ
- Source links
Direct answer
The initiative does not require every agent to adopt one technology immediately. It calls for cooperation around trusted identity, discovery, protocol compatibility, security governance, and data protection as agents begin to work across platforms and organizations.
Fact sources
On July 17, 2026, the World Internet Conference Asia-Pacific Summit released an initiative calling for mutual trust, connectivity, and interoperability among AI agents. Its nine proposals cover ecosystem development, security governance, open-source collaboration, standard interfaces, privacy and data protection, and closing digital divides. A chair's statement published the same day called for international consensus, coordinated standards, and responsible agent development. Earlier, the ITU announced a Focus Group on identity and access management for agentic AI on July 9. The Linux Foundation announced its intent to launch Agent Name Service on June 23, 2026 for neutral naming, discovery, and authenticity checks, while its A2A project supports agent-to-agent communication. These are cooperation and standards-building efforts, not a single mandatory global standard already deployed everywhere.
Definition, scenarios, steps, and risks
Agent trust means verifying who an agent is, whom it represents, what it may do, and whether its capability claims are credible. Connectivity means controlled discovery and connection. Interoperability means exchanging tasks, status, results, and errors through agreed interfaces.
- Define one low-risk, repeatable cross-agent task with explicit inputs, outputs, handoff conditions, and stop conditions.
- Give every agent a separate identity and least-privilege scope; begin with read, discovery, and simulated calls only.
- Pin protocol, tool, and endpoint versions, documenting the role of A2A, MCP, APIs, or custom adapters.
- Verify each agent's name, origin, capability claims, destination, and authorization scope, while logging requests and tool calls.
- Route writes, payments, deletions, account changes, and external messages through explicit human approval.
- Test timeouts, revocation, expired identities, network failure, and rollback, then measure errors, latency, and review cost.
Risks include identity spoofing, false capability claims, prompt injection, excessive permissions, cross-platform data leakage, incompatible protocol versions, and treating a cooperation initiative as an already binding global standard.
Why it matters
When one agent can call another agent, tool, or account, fluent output is no longer enough. Identity, permissions, provenance, auditability, and recovery determine whether cross-agent workflows can enter real operations.
Impact for ordinary AI users
Ordinary users will see more features that delegate work to other agents, connect external tools, and automate multi-step tasks. Before enabling them, check the destination, permission scope, data path, paid actions, external messages, and revocation method.
Related tools/tutorials
Continue through ENHE AI frontier news, software, account services, skill tutorials, and local deployment guidance to move from protocol concepts to permission checks.
Related ENHE AI links: AI frontier news, AI software and local deployment tools, AI account services and permission management, AI skill tutorials and validation methods, ENHE AI homepage.
FAQ
Are A2A and MCP the same protocol?
No. A2A primarily handles task communication and status exchange between agents, while MCP commonly connects models or agents to tools, data, and context. They can be combined.
Is the global agent cooperation initiative already a mandatory standard?
No. It is a cooperation initiative released on July 17, 2026, while related standards, identity infrastructure, and protocol governance continue to develop.
Why should ordinary users care about agent interoperability?
More AI software will connect email, calendars, storage, code, accounts, and other agents, making identity, permissions, logs, revocation, and rollback direct user concerns.
Source links
- 新华社：推动全球智能体互信互联互通合作倡议
- 中华人民共和国外交部：世界互联网大会亚太峰会主席声明
- ITU: Focus Group on Identity and Access Management for Agentic AI
- 中央网信办：构筑智能体创新发展的安全底座
- Linux Foundation: Agent Name Service trusted identity infrastructure
- Linux Foundation: Agent2Agent Protocol Project
What this means for everyday users
ENHE users should add agent identity, protocol compatibility, permission scope, log export, human approval, and revocation to software, account-service, and local-deployment checklists.
Related tutorials
Related reading
From Chat Boxes to Personal AI Companions: AI Assistants Are Entering the Desktop Execution Era
AI assistants are moving from answering questions toward continuing real tasks. AI agents, MCP tool ecosystems, personal memory, and local workbenches are pushing this shift together. For users, the real value is not another chat box, but less repeated context setup and more continuity from thinking to doing.
Meta introduces Muse with a dedicated secure VM, Sentinel checks, and approval gates
Meta introduced the Muse personal AI agent on September 8 and began rolling it out in the United States on iOS, Android, and the web. Muse runs inside a dedicated Secure VM with its own browser and can continue tasks such as planning, form filling, and work across connected applications after the user closes the app. Meta says a system-isolated Sentinel agent reviews every action before it reaches the internet, while sensitive steps such as sending an email or making a purchase require user approval. Users can choose connected services, change access, disconnect them, and inspect an audit trail. These security, privacy, and performance claims come from Meta and should be independently tested with low-risk tasks before broader delegation.
GitHub adds enterprise controls for Copilot agent commands, files, and network access
GitHub released enterprise-managed permissions for Copilot agent operations on September 9. Administrators can centrally set shell commands, file reads and writes, and access to network domains to blocked, approval required, or allowed without a prompt. User preferences, workspace settings, automatic approval, and earlier approvals cannot make the enterprise policy less restrictive. GitHub says the controls are generally available in the Copilot app, Copilot CLI, and Visual Studio Code sessions that use Agent Host for Copilot Business and Enterprise customers. Security and platform teams should begin with a minimum-permission baseline, test representative repositories, and expand only the operations that have a clear owner, audit trail, and rollback path.
An AI agent system-card checklist for purpose, components, evaluation, monitoring, and ownership
The UK Ministry of Defence Digital AI Practitioner's Handbook says a system card should be created when AI models are selected or shortlisted, updated throughout the system lifecycle, and kept with earlier versions to preserve an audit trail. Its guidance calls for a system overview and responsible roles, technical details about models and hosting, intended use and users, operating and training requirements, monitoring plans, and supporting documentation. This article adapts that Defence guidance into a general release evidence card for AI agents, adding prompts, tools, permissions, evaluations, known limits, and recovery paths. Those additions are an ENHE AI engineering interpretation, not a claim that the UK guidance creates a legal requirement for other organizations or defines one universal agent schema.
Mistral reports a 40,000-line Fortran-to-C++ migration built around numerical parity
Mistral published a legacy-modernization case study on September 9 involving a 300,000-line Fortran 77 reservoir simulator for an unnamed European energy operator. The first sprint migrated 40,000 lines of core functionality to C++. Before migration, the team built a numerical-parity harness that compared final outputs and critical intermediate checkpoints, then used more than one hundred agents to document the caller-callee tree. Mistral says a fully autonomous first attempt produced working code that still resembled Fortran written in C++ syntax. The successful approach divided modules into manageable units and coordinated planning, coding, testing, and review, with engineers resolving blocked work. The report supports a practical rule: create a runnable baseline and measurable parity before scaling agent activity.
AWS connects AgentCore evaluations to GitHub Actions for pre-merge agent regression gates
AWS published a reference workflow on September 8 for integrating Amazon Bedrock AgentCore Evaluations with GitHub Actions. The pipeline deploys a development agent and an OAuth-protected MCP server, invokes representative prompts, collects OpenTelemetry traces, and scores behavior before allowing a pull request to proceed. AWS describes built-in dimensions such as helpfulness, correctness, goal success, tool selection, tool parameters, and trajectory order, with custom and code-based evaluators available. The post compares three authentication patterns: evaluating stored traces, using a pre-authorized test user, or issuing machine-to-machine credentials for CI. Its implementation uses the third pattern. Teams should validate that CI credentials cannot escape their intended environment, that evaluation samples cover important failures, and that score variance, latency, evaluator cost, and rollback behavior are visible before treating the threshold as a release gate.
Summary
Trust, connectivity, and interoperability are becoming foundational for deployed AI agents. Useful cross-agent workflows require clear identities, constrained permissions, compatible interfaces, visible evidence, and reversible actions.
