Section 11 · Section
Safety, Security and Alignment
السلامة والأمن والمواءمة
This chapter treats the technical side of safety, security and alignment: how systems are tested, secured, understood and steered toward their intended goals, and how their risks are tracked. It differs from Chapter 12, which handles normative and societal dimensions, from Chapter 14, which presents frameworks and legislation, and from Chapter 9, which concerns measurement and evaluation.
Section11
Branches8 branches
Entries0 entries
§ 11.0
Branches
الفروع| Code | Branch | Scope | Entries |
|---|---|---|---|
| 11.1 | Assurance, TEVV, Audit and Certification | Covers the technical assurance regime: testing, verification, validation, audit and certification as practices that secure a system's reliability in its context | 0 |
| 11.2 | Cybersecurity and Adversarial Machine Learning | Covers the security of artificial intelligence systems and defense against attacks on them: prompt injection, data poisoning, model theft, adversarial examples and software supply-chain security | 0 |
| 11.3 | Physical Safety of Robots, Vehicles and Medical Devices | Covers the safety of systems that act in the physical world and may cause bodily harm: robots, autonomous vehicles and medical devices, together with the safety engineering and verification that go with them | 0 |
| 11.4 | Interpretability and Model Reverse-Engineering | Covers understanding what happens inside models: explaining decisions, analysing internal representations and circuits, methods for reverse-engineering network behaviour, and their limits | 0 |
| 11.5 | Deepfakes, Authenticity and Watermarking | Covers misleading machine-generated content, methods for detecting it and proving the authenticity of media, and technical watermarks and provenance data | 0 |
| 11.6 | Technical and Value Alignment | Covers the problem of making systems behave in line with human goals and values: scalable oversight, reward and control, goal drift and value alignment | 0 |
| 11.7 | Incidents and Systemic Risks | Covers the classification of system failures, their harms and the wide-ranging risks to society and critical infrastructure: analytical methods, failure patterns and systemic scenarios | 0 |
| 11.8 | Dual Use and Offensive Uses | Covers capabilities that can serve both good and harm and their potential for deliberate abuse: chemical, biological, cyber and manipulation risks, and technical ways to limit them | 0 |