2026负责任人工智能（AI）尽职调查指南_61页_2mb

OECD Due Diligence Guidance for Responsible AI Summary

Core Content

The OECD Due Diligence Guidance for Responsible AI is a comprehensive framework designed to help multinational enterprises implement responsible AI practices in alignment with the OECD Guidelines for Multinational Enterprises (MNE Guidelines) and the OECD Recommendation on Artificial Intelligence (AI Principles). It outlines a structured approach for identifying, addressing, and mitigating risks associated with AI systems throughout their lifecycle.

The guidance is structured around a six-step due diligence framework, which includes embedding responsible business conduct (RBC) into policies, identifying and assessing risks, ceasing or mitigating adverse impacts, tracking implementation, communicating actions, and cooperating in remediation. These steps are supported by practical examples and a roadmap of related provisions from existing AI risk management frameworks.

The document emphasizes the importance of stakeholder engagement, particularly workers, in the development and use of AI systems. It advocates for AI to be seen as an enhancement to human capability rather than a replacement for human labor. This approach ensures that AI systems are developed and used in ways that support ethical, sustainable, and socially responsible outcomes.

Key Considerations

• AI as a Transformative Technology: AI has the potential to significantly enhance productivity, create economic value, and solve complex challenges across various sectors such as healthcare, manufacturing, logistics, and public administration.
• RBC as a Holistic Approach: Responsible AI development and use require a "whole-of-value chain" approach, considering not only the technological innovations but also the role of data suppliers, financial inputs, and physical infrastructure.
• Risk-Based Approach: The guidance promotes a risk-based method for due diligence, ensuring that enterprises proactively identify and address both actual and potential adverse impacts of AI systems.
• Stakeholder Engagement: Enterprises are encouraged to engage with workers and other stakeholders to ensure AI systems are developed and used in a way that complements human capabilities and addresses societal concerns.

Target Audience

The primary audience of the guidance includes:

• Group 1: Suppliers of AI Inputs: These enterprises provide data, code, algorithms, and other resources necessary for AI system development. They include data annotation services, compute providers, cloud service providers, and hardware manufacturers.
• Group 2: Enterprises Active in the AI System Lifecycle: These enterprises are involved in the planning, design, development, testing, deployment, and operation of AI systems. They may also modify and re-deploy existing models for specific use cases.
• Group 3: Users of AI Systems: These enterprises utilize AI systems in their operations, products, and services. They include financial institutions and companies in the real economy, such as manufacturers and sellers of goods and services.

The guidance also addresses SMEs, acknowledging their unique challenges in implementing RBC due diligence. It encourages SMEs to engage in collaborative approaches and industry initiatives to reduce costs and improve compliance with international standards.

Main Views

• Responsible AI is a Business Imperative: Enterprises that demonstrate commitment to responsible AI development can gain competitive advantages, access capital markets, and attract premium business relationships.
• Global Regulatory Alignment: The guidance supports international cooperation and policy coherence by aligning with existing AI risk management frameworks and promoting interoperability.
• Adaptability and Proportionality: The nature and extent of due diligence should be proportional to the size of the enterprise, its involvement with adverse impacts, and the severity of those impacts.
• Transparency and Accountability: The guidance emphasizes the importance of transparency, explainability, and traceability in AI systems to ensure accountability and trustworthiness.

Key Information

• Framework Structure: The guidance is based on a six-step due diligence framework:

  1. Embed RBC into policies and management systems.
  2. Identify and assess actual and potential adverse impacts.
  3. Cease, prevent, and mitigate adverse impacts.
  4. Track implementation and results.
  5. Communicate actions to address impacts.
  6. Provide for or cooperate in remediation.

• Practical Implementation: Each step includes practical examples and a roadmap of related provisions from other AI risk management frameworks, helping enterprises understand how to apply the guidance in their specific contexts.

• Supporting Mechanisms: The guidance includes boxes and notes that offer insights into specific considerations for SMEs, the role of National Contact Points (NCPs), and the importance of collaboration in addressing AI-related risks.

• Legal and Ethical Considerations: The guidance acknowledges the need for enterprises to consider the legal and ethical implications of AI systems, particularly in contexts where laws may not align with international RBC standards.

• Global Market Access: Responsible AI is becoming a crucial factor in accessing global markets. Enterprises that integrate harm prevention into their AI development and adoption processes can position themselves advantageously for cross-border expansion.

Conclusion

The OECD Due Diligence Guidance for Responsible AI is a vital resource for enterprises seeking to implement responsible AI practices in a global context. It supports innovation, investment, and growth by providing a clear and structured approach to managing AI-related risks and promoting ethical, sustainable, and socially responsible AI development. The guidance encourages a collaborative and proportionate approach to due diligence, ensuring that all stakeholders, including SMEs and workers, are considered in the development and use of AI systems.