“Requirements of Trustworthy AI The principles outlined in Chapter I must be translated into concrete requirements to achieve Trustworthy AI. These requirements are applicable to different stakeholders partaking in AI systems’ life cycle: developers, deployers and end-users, as well as the broader society. By developers, we refer to those who research, design and/or develop AI systems. By deployers, we refer to public or private organisations that use AI systems within their business processes and to offer products and services to others. End-users are those engaging with the AI system, directly or indirectly. Finally, the broader society encompasses all others that are directly or indirectly affected by AI systems. Different groups of stakeholders have different roles to play in ensuring that the requirements are met: a. Developers should implement and apply the requirements to design and development processes; b. Deployers should ensure that the systems they use and the products and services they offer meet the requirements; c. End-users and the broader society should be informed about these requirements and able to request that they are upheld. The below list of requirements is non-exhaustive.35 It includes systemic, individual and societal aspects: 1 Human agency and oversight Including fundamental rights, human agency and human oversight 2 Technical robustness and safety Including resilience to attack and security, fall back plan and general safety, accuracy, reliability and reproducibility 3 Privacy and data governance Including respect for privacy, quality and integrity of data, and access to data 4 Transparency Including traceability, explainability and communication 5 Diversity, non-discrimination and fairness Including the avoidance of unfair bias, accessibility and universal design, and stakeholder participation 6 Societal and environmental wellbeing Including sustainability and environmental friendliness, social impact, society and democracy 7 Accountability Including auditability, minimisation and reporting of negative impact, trade-offs and redress.” (High-Level Expert Group on AI, 2019, p. 14)“While all requirements are of equal importance, context and potential tensions between them will need to be taken into account when applying them across different domains and industries. Implementation of these requirements should occur throughout an AI system’s entire life cycle and depends on the specific application. While most requirements apply to all AI systems, special attention is given to those directly or indirectly affecting individuals. Therefore, for some applications (for instance in industrial settings), they may be of lesser relevance. The above requirements include elements that are in some cases already reflected in existing laws. We reiterate that – in line with Trustworthy AI’s first component – it is the responsibility of AI practitioners to ensure that they comply with their legal obligations, both as regards horizontally applicable rules as well as domain-specific regulation” (High-Level Expert Group on AI, 2019, p. 15)The full document expands on each in detail.“To implement the above requirements, both technical and non-technical methods can be employed. These encompass all stages of an AI system’s life cycle. An evaluation of the methods employed to implement the requirements, as well as reporting and justifying51 changes to the implementation processes, should occur on an ongoing basis. AI systems are continuously evolving and acting in a dynamic environment.” (High-Level Expert Group on AI, 2019, p. 20)