欧洲议会-人工智能的民事责任制度（英文）-2020.9-232页

Summary of the Document: Civil Liability Regime for Artificial Intelligence

Core Content

This document, Civil liability regime for artificial intelligence, presents a comprehensive analysis of the potential added value of establishing a unified EU civil liability framework for artificial intelligence (AI) systems. It outlines the current regulatory landscape, the socio-economic functions of liability rules in the AI context, and proposes policy options for a coherent EU approach.

The report is part of the European Added Value Assessment (EAVA) and aims to support the European Parliament in its legislative initiatives regarding AI liability. It is authored by Dr. Tatjana Evas and supported by legal experts from across the EU Member States.

---

Main Socio-Economic Functions of Civil Liability Rules in the AI Context

1. Incentive for Rational Resource Use and Fair Risk Distribution

   • Liability rules encourage responsible use of AI technologies and ensure fair distribution of risks among stakeholders.
   • They aim to reflect justice, equity, and fairness in how risks are allocated.

2. Encouragement of Innovation and Reduction of Uncertainty Costs

   • Clear liability rules reduce uncertainty for economic actors, which can promote innovation investment.
   • Ex-ante regulation can mitigate the costs associated with legal uncertainty and litigation.

3. Enhancement of Consumer Trust and Acceleration of AI Diffusion

   • A transparent liability framework increases consumer confidence in AI technologies.
   • This can lead to faster adoption and integration of AI across various industries.

---

Current EU Civil Liability Framework for AI

• The EU currently lacks a specific civil liability regime for AI, relying instead on the EU Product Liability Directive (PLD), which applies to defective products.
• The PLD provides a no-fault liability regime, meaning producers can be held liable for damage caused by defective products, regardless of fault.
• However, it is unclear whether the PLD applies to AI systems, as AI presents unique characteristics such as complexity, autonomy, and potential for unintended consequences.

---

National Liability Regimes and Their Application to AI

• National liability regimes show significant divergence in how they approach AI-related liability.
• Some countries apply strict liability for "things" (e.g., products), while others use fault-based liability or dangerous activities frameworks.
• There are also sector-specific regulations, particularly for drones, autonomous vehicles, and medical AI.
• These differences may lead to regulatory fragmentation, creating obstacles for the internal market and increasing legal and operational costs for businesses.

---

Policy Options and Approaches for EU-Level AI Liability Regulation

• The document explores two main liability approaches:
  • Strict liability
  • Fault-based liability
• It argues that a common EU approach is necessary to ensure consistency, reduce uncertainty, and promote the safe and widespread adoption of AI.
• The European Parliament has proposed various policy initiatives, including a 2017 resolution and a 2020 legislative draft report.
• The European Commission has also engaged in policy discussions, including expert reports on liability and new technologies.
• The report suggests that insurance schemes (obligatory or voluntary) could be introduced to support the liability regime.

---

Quantitative Assessment of Added Value

• A quantitative analysis suggests that EU action on AI liability could generate €54.8 billion in added value by 2030 through increased R&D investment.
• If broader impacts such as reduced accidents, health and environmental benefits, and user impacts are considered, the added value could reach €498.3 billion.
• The assessment is based on data from previous studies, including the Cost of Non-Europe and European Added Value Assessment on Autonomous Vehicles.
• These studies indicate that clear liability rules can lead to increased innovation, investment, and consumer trust, ultimately enhancing the competitiveness of the EU economy.

---

Conclusion

• A clear and coherent EU civil liability regime for AI is essential to reduce risks, increase safety, and foster innovation and trust.
• The current lack of harmonization leads to regulatory fragmentation, which may hinder the development and diffusion of AI.
• Timely and coordinated EU action on AI liability is necessary to maximize economic and social added value and ensure a fair and effective risk distribution mechanism.
• The potential benefits of a unified regime include reduced legal costs, enhanced consumer rights, and increased business confidence.

---

Key Recommendations

• Establish a common EU liability framework for AI systems.
• Ensure legal clarity and coherence to support innovation and investment.
• Consider the social and economic impacts of liability rules in shaping policy.
• Introduce insurance mechanisms to support the liability regime.
• Align EU liability rules with existing national systems to avoid fragmentation and ensure market efficiency.

---

Added Value

• The economic benefits of a unified liability regime include increased R&D, reduced accidents, and improved consumer trust.
• The social benefits involve fair risk distribution and enhanced consumer protection.
• The potential added value is estimated at €54.8 billion to €498.3 billion by 2030, depending on the scope of the analysis.
• A coherent liability regime would boost the competitiveness of the EU economy and support the safe and responsible diffusion of AI technologies.

---

Annexes

• Annex I: Comparative study on national rules concerning non-contractual liability, including AI.
• Annex II: Synopsis of non-material damage in national law.
• The annexes provide detailed legal comparisons across 19 Member States and outline the different approaches to non-material damage.

---

Tables and Figures

• Table 1: Main points of discussion relating to the application of the PLD to AI.
• Table 2–13: Comparative analysis of strict liability systems across Member States.
• Table 14–31: Overview of liability regimes and their economic impacts.
• Figure 1: How the scenarios are modelled in the quantitative assessment.

---

Final Note

This report underscores the importance of a timely and coherent EU civil liability regime for AI to ensure economic growth, consumer trust, and regulatory consistency. It serves as a foundational document for future legislative and policy decisions on AI liability.