10 | Conclusions | 0 |
5.2 | Recognition of layers of trust | 1 |
5.3 | Application of software and data quality standards | 0 |
5.4 | Application of risk management | 4 |
5.5 | Hardware-assisted approaches | 0 |
6.1 | General concepts | 26 |
6.2 | Types of stakeholders | 0 |
6.3 | Assets | 0 |
6.4 | Values | 0 |
7.1 | Responsibility, accountability and governance | 0 |
7.2 | Safety | 0 |
8.1 | General: vulnerabilities, threats and challenges | 3 |
8.10 | System hardware faults | 1 |
8.2 | AI specific security threats | 0 |
8.2.1 | AI specific security threats: general | 0 |
8.2.2 | Data poisoning | 2 |
8.2.3 | Adversarial attacks | 0 |
8.2.4 | Model stealing | 0 |
8.2.5 | Hardware-focused threats to confidentiality and integrity | 0 |
8.3 | AI specific privacy threats | 0 |
8.3.1 | AI specific privacy threats: general | 3 |
8.3.2 | Data acquisition | 3 |
8.3.3 | Data pre-processing and modelling | 1 |
8.3.4 | Model query | 0 |
8.4 | Bias | 1 |
8.5 | Unpredictability | 1 |
8.6 | Opaqueness | 0 |
8.7 | Challenges related to the specification of AI systems | 0 |
8.8 | Challenges related to the implementation of AI systems | 0 |
8.8.1 | Data acquisition and preparation | 1 |
8.8.2 | Modelling | 0 |
8.8.3 | Model updates | 0 |
8.8.4 | Software defects | 0 |
8.9 | Challenges related to the use of AI systems | 0 |
8.9.1 | Human-computer interaction (HCI) factors | 0 |
8.9.2 | Misapplication of AI systems that demonstrate realistic human behaviour | 0 |
9.1 | General: mitigation measures | 3 |
9.10 | Testing and evaluation | 1 |
9.10.1 | Testing and evaluation: general | 3 |
9.10.2 | Software validation and verification methods | 0 |
9.10.3 | Robustness considerations | 0 |
9.10.4 | Privacy-related considerations | 0 |
9.10.5 | System predictability considerations | 0 |
9.11 | Use and applicability | 0 |
9.11.1 | Compliance | 5 |
9.11.2 | Managing expectations | 0 |
9.11.3 | Product labelling | 0 |
9.11.4 | Cognitive science research | 0 |
9.2 | Transparency | 10 |
9.3 | Explainability | 1 |
9.3.1 | Explainability: general | 3 |
9.3.2 | Aims of explanation | 1 |
9.3.3 | Ex-ante vs ex-post explanation | 0 |
9.3.4 | Approaches to explainability | 0 |
9.3.5 | Modes of ex-post explanation | 1 |
9.3.6 | Levels of explainability | 0 |
9.3.7 | Evaluation of the explanations | 2 |
9.4 | Controllability | 1 |
9.4.1 | Controllability: general | 3 |
9.4.2 | Human-in-the-loop control points | 0 |
9.5 | Strategies for reducing bias | 0 |
9.6 | Privacy | 4 |
9.7 | Reliability, resilience and robustness | 1 |
9.8 | Mitigating system hardware faults | 0 |
9.9 | Functional safety | 1 |