EU Policy and Legal Framework for Artificial Intelligence, Robotics and Related Technologies - The AI Act

Description

Artificial Intelligence (AI) can benefit our society and economy, but also brings with it new challenges and raises legal and ethical questions. According to the author of this comprehensive analysis, it is imperative to ensure that AI is developed and applied in an appropriate legal and regulatory framework that promotes innovation and investment and, at the same time, addresses the risks associated with certain uses of AI-related technologies. Essential to understanding the relationship between policy and law, this book traces the evolution of EU policy on artificial intelligence and robotics, focusing in particular on the EU’s ethical framework for AI, which defines trust as a prerequisite for ensuring a human-centric approach. The main part of the book provides a thorough and systematic analysis of the Commission’s 2021 proposed AI Act, which establishes harmonised rules for the development, placement on the market and use of AI systems in the EU. The author painstakingly compares the Commission’s proposed AI Act with the numerous “compromise” proposals of the Council of the European Union, leading to the final version of the Council’s AI Act (general approach) and its formal adoption on 6 December 2022. The author also examines with extraordinary detail the amendments proposed by the relevant committees and political groups of the European Parliament, revealing the position the Parliament is likely to adopt in the forthcoming negotiations with the Commission and the Council on the text of the AI Act. Numerous legislative and policy documents are presented in detail, while the analysis also considers the comments made by all interested parties (e.g. the European Commission, Council of the European Union, European Parliament, governmental organisations, national competent authorities, and stakeholders/actors with different/conflicting interests, such as corporations, business and consumer associations, civil society and other non-profit organisations).  In the course of its in-depth analysis, this book will provide readers with crucial insight into the reasons behind the European Institutions’ different approaches and the often contradictory interests of stakeholders. Because the policy arguments are carefully balanced and drafted with scrupulous care, this volume will establish itself as a reference resource to be consulted for years to come. Contents Part I: Introduction: The Evolution of EU Policy and Law for Artificial Intelligence-Launching a European Initiative on Artifi... Chapter 1: Introduction: Benefits/Opportunities and Threats/Risks of AI-Structure of this Book Legal, Policy Documents and Other Sources Chapter 2: Launching a European Initiative on Artificial Intelligence 2.1 The European Parliament Resolution of February 2017 with Recommendations to the Commission on Civil Law Rules and Ethical ... 2.2 European Commission´s Follow Up to the 2017 European Parliament Resolution 2.3 European Parliament: Public Consultation on the Future of Robotics and Artificial Intelligence 2.4 The European Council of October 2017 2.5 Declaration of Cooperation on Artificial Intelligence 2.6 Commission´s 2018 Communication on Artificial Intelligence for Europe 2.6.1 Introduction 2.6.2 Boosting the EU´s Technological and Industrial Capacity and AI Uptake Across the Economy 2.6.3 Preparing for Socioeconomic Changes Brought About by Artificial Intelligence 2.6.4 Ensuring an Appropriate Ethical and Legal Framework Draft AI Ethics Guidelines Safety and Liability Empowering Individuals and Consumers to Make the Most of AI Summary of Key Actions to be Taken by the Commission to Ensure an Appropriate Legal Framework for AI 2.6.5 Joining Forces Engaging Members States and Stakeholders: A Coordinated Plan on AI and the European AI Alliance Monitoring AI Development and International Outreach 2.6.6 Conclusion 2.7 The EU´s 2018 Coordinated Plan on Artificial Intelligence 2.7.1 Introduction 2.7.2 Main Objectives and Initiatives of the EU´s Coordinated Plan on AI Common Objectives and Complementary Efforts Towards a European AI Public-Private Partnership: Maximising Investments Through Partnerships From the Lab to the Market: Strengthening Research Excellence in Trustworthy AI Technologies and Broad Diffusion Adapting Skills, Learning and Training Programmes to Better Prepare for AI Data Is a Cornerstone for AI: Creating a Common European Data Space-Public Sector Ethics by Design, Innovation Friendly Legal Framework, Security- Related Aspects of AI Applications and International Cooperat... 2.8 The Council of the European Union of 18 February 2019 2.9 Conclusion of Part I Legal, Policy Documents and Other Sources Part II: The EU Ethical Framework for Trustworthy Artificial Intelligence Chapter 3: Ethical Principles for Trustworthy AI 3.1 Overview of the European AI Strategy 3.2 Building Trust in Human-Centric Artificial Intelligence 3.3 Notion of Human-Centric and Trustworthy AI 3.4 A Framework for Trustworthy AI-Three Components: Lawful, Ethical, Robust AI 3.5 Foundations of Trustworthy AI: Fundamental Rights and Ethical Principles for Ethical and Robust AI 3.6 Fundamental Rights as a Basis for Trustworthy AI 3.6.1 Respect for Human Dignity 3.6.2 Freedom of the Individual 3.6.3 Respect for Democracy, Justice and the Rule of Law: Citizens´ Rights 3.6.4 Equality, Non-Discrimination and Solidarity: Including the Rights of Persons at Risk of Exclusion 3.7 Ethical Principles in the Context of AI Systems: The Ethical Principles That Must Be Adhered to in Order to Ensure Ethical... 3.7.1 European Group on Ethics in Science and New Technologies (``EGE´´): Statement on ``Artificial Intelligence, Robotics and... 3.7.2 The AI4People´s Scientific Committee 3.7.3 The Four Ethical Principles Identified by HLEG´s Ethics Guidelines for Trustworthy AI (i) The Principle of Respect for Human Autonomy (ii) The Principle of Prevention of Harm (Non-Maleficence) (iii) The Principle of Fairness/Justice (iv) The Principle of Explicability (v) The Principle of Beneficence (`Do Only Good´) 3.7.4 Tensions Between the Principles and Trade-Offs: Impact-Based Approach Legal, Policy Documents and Other Sources References Chapter 4: Realising Trustworthy AI: Implementation of the Key Ethical Requirements 4.1 Realising Trustworthy AI-Seven Key Ethical Requirements 4.1.1 Human Agency and Oversight 4.1.2 Technical Robustness and Safety 4.1.3 Privacy and Data Governance 4.1.4 Transparency 4.1.5 Diversity, Non-Discrimination and Fairness 4.1.6 Societal and Environmental Well-Being 4.1.7 Accountability 4.2 Ensuring the Implementation of the Seven Key Requirements-Technical and Non-technical Methods to Realise Trustworthy AI 4.3 Assessing Trustworthy AI-Adopting an Assessment List 4.3.1 Fundamental Rights Impact Assessment (FRIA) 4.3.2 Requirement #1: Human Agency and Oversight 4.3.3 Requirement #2: Technical Robustness and Safety 4.3.4 Requirement #3: Privacy and Data Governance 4.3.5 Requirement #4: Transparency 4.3.6 Requirement #5: Diversity, Non-discrimination and Fairness 4.3.7 Requirement #6: Societal and Environmental Well-being 4.3.8 Requirement #7: Accountability 4.4 Conclusion of Part II (Chaps. 3 and 4) Legal, Policy Documents and Other Sources References Part III: Shifting from a Soft-Law Approach Towards a Legislative Process: Setting Out Options on How to Address the Risks for... Chapter 5: A European Approach to Excellence and Trust: The 2020 White Paper on Artificial Intelligence 5.1 Introduction 5.2 Ecosystem of Excellence: The Policy Framework for AI 5.3 Ecosystem of Trust: The Regulatory Framework for AI 5.3.1 Introduction-Risks for Fundamental Rights, Safety, and Liability 5.3.2 Adjustments to Existing EU Legislative Framework Relating to AI: Issues to Be Addressed Effective Application and Enforcement of Existing EU and National Legislation Limitations of Scope of Existing EU Legislation Changing Functionality of AI Systems Uncertainty as Regards the Allocation of Responsibilities Between Different Economic Operators in the Supply Chain Changes to the Concept of Safety 5.3.3 Scope of a Future EU Regulatory Framework on AI Definition of AI Scope of Application: Risk-Based Approach-``High-Risk´´ Sectors and ``High-Risk´´ AI Applications Mandatory Legal Requirements to Be Imposed on Relevant Actors for High-Risk AI Applications Economic Operators and Geographic Scope of the Legal Requirements to Be Applied in Relation to the High-Risk AI Applications Compliance and Enforcement-Conformity Assessment Voluntary Labelling for No-High Risk AI Applications Governance 5.3.4 Conclusion and the Commission´s Inception Impact Assessment Legal, Policy Documents and Other Sources Reference Chapter 6: The European Parliament´s 2020 Resolution: Proposal for a Regulation on Ethical Principles for the Development, Dep... 6.1 Introduction 6.2 Purpose and Scope of the Parliament´s Proposed Regulation 6.3 Principles of the Parliament´s Proposed Regulation: Obligations for High-Risk Technologies 6.3.1 Human-centric, Human-Made, and Human-Controlled AI 6.3.2 Safety and Transparency 6.3.3 Accountability 6.3.4 Non-Bias and Non-Discrimination 6.3.5 Social Responsibility and Gender Equality 6.3.6 Environmental Sustainability 6.3.7 Respect for Privacy and Protection of Personal Data-Limitations on the Use of Biometric Recognition 6.3.8 Right to Redress 6.3.9 Risk Assessment 6.3.10 Compliance Assessment 6.3.11 European Certificate of Ethical Compliance 6.4 Institutional Oversight 6.4.1 Good Governance and Standards 6.4.2 Supervisory Authorities and Coordination 6.5 Conclusion of Part III (Chaps. 5 and 6) Legal, Policy Documents and Other Sources Part IV: The Emerging EU Legal Framework for Artificial Intelligence Chapter 7: Boosting Excellence in AI: The 2021 Coordinated Plan with Member States 7.1 Introduction 7.2 Four Key Policy Objectives Legal, Policy Documents and Other Sources Chapter 8: The Proposed Artificial Intelligence Act and Subsequent `Compromise´ Proposals: Commission, Council, Parliament 8.1 Introduction: Reasons for and Objectives of the Proposed AI Act 8.2 Subject Matter, Material and Territorial Scope of the Proposed AI Act 8.3 Definition of (Trustworthy) Artificial Intelligence 8.4 Risk-Based and Proportionate Approach 8.5 Highest (Unacceptable) Risk Level: Prohibited Artificial Intelligence Practices 8.5.1 AI Systems Used for Adverse Behavioural Influencing: AI Systems That Distort Human Behaviour and Exploit the Vulnerabili... 8.5.2 AI Systems Providing Social Scoring 8.5.3 Use of AI-Powered Predictive Policing Systems? 8.5.4 Use of Real Time Biometric Identification Systems in Publicly Accessible Spaces for the Purpose of Law Enforcement The Approach Followed by the Commission, the Council´s Compromise Texts, and the Parliament´s Committees The European Parliament´s Position: Resolution of 6 October 2021 8.6 High-Risk AI Systems: Classification of AI Systems as High-Risk 8.6.1 High-Risk AI Systems Are Limited to Those That Have a Significant Harmful Impact on the Health, Safety and Fundamental R... 8.6.2 Categories of High-Risk AI Systems Adapting Existing Product Safety Legislation: Products Which Are Already Subject to a Third-Party Prior Conformity Assessment-... List of High-Risk Use Cases: Stand-Alone AI Systems-Annex III of the AI Act 8.7 Requirements for High-Risk AI Systems 8.7.1 Risk Management System 8.7.2 Data Governance Requirements: Use of High-Quality Data Sets 8.7.3 Technical Documentation 8.7.4 Record-Keeping/Logging Capabilities (Traceability) 8.7.5 Transparency and Provision of Information to Users 8.7.6 Human Oversight 8.7.7 Accuracy, Robustness and Cybersecurity 8.8 Harmonised Standards, Common Specifications, Cybersecurity Requirements: Presumption of Conformity 8.8.1 Harmonised Standards and Common Specifications: Presumption of Conformity 8.8.2 Presumption of Conformity with Certain Requirements, Including Cybersecurity Requirements 8.9 Obligations of Providers and Users of High-Risk AI Systems and Other Parties 8.9.1 Need to Reflect the Complexity of the AI Ecosystem in the Balance of Obligations for Different Roles and Stakeholders Ac... 8.9.2 General Purpose AI Systems 8.9.3 Obligations Imposed on Providers of High-Risk AI Systems Obligation to Establish a Quality Management System Obligation to Draw Up Technical Documentation/Documentation Keeping Conformity Assessment EU Declaration of Conformity CE Marking of Conformity Registration: EU Database for High-Risk AI Systems Listed in Annex III Automatically Generated Logs Cooperation with Competent Authorities, the AI Board and the Commission-Document Retention/Keeping Corrective Actions Post-market Monitoring System and Post-marketing Plan Duty of Information-Reporting of Serious Incidents and of Malfunctioning Appointing an Authorised Representative in the EU 8.9.4 Obligations of Authorised Representatives, Importers, Distributors, Product Manufacturers, Other Third Parties, and User... Conditions for Other Persons to be Subject to the Obligations Imposed on the Provider Obligations of Authorised Representatives Obligations of Importers Obligations of Distributors Obligations of Users of High-Risk AI Systems 8.10 Light-Touch Obligations: Transparency Obligations and Codes of Conduct for Certain (Low-Risk or No-Risk) AI Systems 8.10.1 Transparency Obligations 8.10.2 Codes of Conduct for Voluntary Application of Specific Requirements 8.11 Governance 8.11.1 European Artificial Intelligence Board: Structure and Tasks 8.11.2 National Competent Authorities-National Supervisory Authority 8.11.3 Conformity Assessment, Notifying Authorities, Notified Bodies Ex-ante Conformity Assessment Derogation from Conformity Assessment Procedure Notifying Authorities Conformity Assessment Bodies Applying for Notification-Notification Procedure Notified Bodies Identification Numbers and Lists of Notified Bodies Changes to Notifications Challenge to the Competence of Notified Bodies Coordination of Notified Bodies and Conformity Assessment Bodies of Third Countries 8.12 Market Surveillance and Enforcement 8.12.1 Introduction 8.12.2 Access to Data and Documentation 8.12.3 Procedure for Dealing with AI Systems Presenting a Risk at National Level 8.12.4 EU Safeguard Procedure 8.12.5 Compliant AI Systems Which Present a Risk 8.12.6 Formal Non-compliance 8.12.7 Specific Rules Regarding Enforcement at EU Level-Setting Out the Conditions for Commission Intervention in Enforcing th... Commission´s Intervention and Opening of Proceedings Commission´s Investigation and Enforcement Power 8.12.8 Lodging Complaints Directly with the Relevant Authorities? 8.13 Penalties and Fines-Sanctions 8.14 Measures in Support of Innovation 8.14.1 AI Regulatory Sandboxes 8.14.2 Processing of Personal Data for Developing Certain AI Systems in the Public Interest in the AI Regulatory Sandboxes 8.14.3 Testing of High-Risk AI Systems in Real World Conditions? 8.14.4 Support Measures for Operators, in Particular SMEs, Including Start-Ups 8.15 Delegated Acts 8.16 Latest Developments: Conclusion of This Chapter 8.16.1 The Council of the European Union (``the Council´´)-Process of Preparing the Council´s Compromise Proposal 8.16.2 The European Parliament (``the Parliament´´)-Process of Preparing the Parliament´s Compromise Proposal 8.16.3 Conclusion Legal and Policy Documents Studies and Reports Other Sources References