人工智能侵權中的責任主體及歸責原則

人工智能的自主決策和系統互聯等技術特徵給人工智能侵權責任的認定和分配帶來挑戰。從侵權責任的補償和預防功能出發，規則的設計要考慮不同行為主體對風險的控制力和預防能力，實現技術發展與風險控制的平衡。人工智能本身不宜作為獨立責任主體，原則上應當將人工智能侵權中的責任更多地分配給提供者而非使用者。人工智能屬 產品，應當在動態的投入市場時間段，運用體現了風險分級的“理性人工智能”標準判斷其是否存在缺陷。人工智能提供者承擔產品責任，而不宜承擔不考慮缺陷的更嚴格的無過錯責任。人工智能使用者對人工智能的控制力有限且很大程度上依賴於提供者，在人工智能侵權中應當承擔過錯責任。結合人工智能技術特點判斷使用者注意義務的成立，靈活而有區別地確定注意義務的內容，能夠實現人工智能風險事先規制法與事後救濟法之間的聯動。

The autonomous decision-making and system interconnectivity of artificial intelligence (AI) introduce challenges in determining and allocating liability for AI-related infringements. From the perspective of tort law's compensatory and preventive functions, the design of liability rules must consider the degree of risk control and preventive capabilities of different actors to balance technological development with risk management. AI itself should not be regarded as an independent subject of liability. As a principle, liability in AI-related infringements should be primarily allocated to AI providers rather than users. Given that AI is a product, the assessment of whether it is defective should be based on a rational AI standard that incorporates a risk-tiered approach throughout the dynamic period of market deployment. AI providers should bear product liability but should not be subject to an excessively strict no-fault liability that disregards the existence of defects. AI users, on the other hand, have limited control over AI and largely rely on providers. Therefore, their liability in AI-related infringements should be based on fault. To establish users' duty of care, it is essential to assess this duty in light of AI's technical characteristics and flexibly differentiate its content. This approach enables an effective interaction between exante regulatory mechanisms and ex-post remedial measures in AI risk governance.