Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AI Hazard Management: A framework for the systematic management of root causes for AI risks

Ronald Schnitzer, Andreas Hapfelmeier, Sven Gaube, Sonja Zillner

TL;DR

This paper presents the AI Hazard Management (AIHM) framework for systematic management of AI risks by identifying root causes (AI hazards) across the AI lifecycle. It defines a three-axis AI hazard taxonomy (life cycle stage, hazard mode, hazard level) and a comprehensive AI hazard list (AIH 1–24) to enable structured hazard identification, risk assessment, and risk treatment, with a strong emphasis on auditability through documented evidence. The framework is designed to run in parallel with AI development, aligning with ISO 31000 risk management principles, and is demonstrated through a power-grid case study using a DNN classifier to detect high-impedance ground faults, where four hazards are analyzed and mitigated. The work lays out a path for broader deployment, including extending the hazard list and developing quantification and mitigation techniques, to improve safety and compliance in AI-enabled systems across domains.

Abstract

Recent advancements in the field of Artificial Intelligence (AI) establish the basis to address challenging tasks. However, with the integration of AI, new risks arise. Therefore, to benefit from its advantages, it is essential to adequately handle the risks associated with AI. Existing risk management processes in related fields, such as software systems, need to sufficiently consider the specifics of AI. A key challenge is to systematically and transparently identify and address AI risks' root causes - also called AI hazards. This paper introduces the AI Hazard Management (AIHM) framework, which provides a structured process to systematically identify, assess, and treat AI hazards. The proposed process is conducted in parallel with the development to ensure that any AI hazard is captured at the earliest possible stage of the AI system's life cycle. In addition, to ensure the AI system's auditability, the proposed framework systematically documents evidence that the potential impact of identified AI hazards could be reduced to a tolerable level. The framework builds upon an AI hazard list from a comprehensive state-of-the-art analysis. Also, we provide a taxonomy that supports the optimal treatment of the identified AI hazards. Additionally, we illustrate how the AIHM framework can increase the overall quality of a power grid AI use case by systematically reducing the impact of identified hazards to an acceptable level.

AI Hazard Management: A framework for the systematic management of root causes for AI risks

TL;DR

This paper presents the AI Hazard Management (AIHM) framework for systematic management of AI risks by identifying root causes (AI hazards) across the AI lifecycle. It defines a three-axis AI hazard taxonomy (life cycle stage, hazard mode, hazard level) and a comprehensive AI hazard list (AIH 1–24) to enable structured hazard identification, risk assessment, and risk treatment, with a strong emphasis on auditability through documented evidence. The framework is designed to run in parallel with AI development, aligning with ISO 31000 risk management principles, and is demonstrated through a power-grid case study using a DNN classifier to detect high-impedance ground faults, where four hazards are analyzed and mitigated. The work lays out a path for broader deployment, including extending the hazard list and developing quantification and mitigation techniques, to improve safety and compliance in AI-enabled systems across domains.

Abstract

Recent advancements in the field of Artificial Intelligence (AI) establish the basis to address challenging tasks. However, with the integration of AI, new risks arise. Therefore, to benefit from its advantages, it is essential to adequately handle the risks associated with AI. Existing risk management processes in related fields, such as software systems, need to sufficiently consider the specifics of AI. A key challenge is to systematically and transparently identify and address AI risks' root causes - also called AI hazards. This paper introduces the AI Hazard Management (AIHM) framework, which provides a structured process to systematically identify, assess, and treat AI hazards. The proposed process is conducted in parallel with the development to ensure that any AI hazard is captured at the earliest possible stage of the AI system's life cycle. In addition, to ensure the AI system's auditability, the proposed framework systematically documents evidence that the potential impact of identified AI hazards could be reduced to a tolerable level. The framework builds upon an AI hazard list from a comprehensive state-of-the-art analysis. Also, we provide a taxonomy that supports the optimal treatment of the identified AI hazards. Additionally, we illustrate how the AIHM framework can increase the overall quality of a power grid AI use case by systematically reducing the impact of identified hazards to an acceptable level.
Paper Structure (14 sections, 2 figures, 1 table)

This paper contains 14 sections, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Definitions and background information
  4. AI Hazard Taxonomy and List
  5. AI hazard taxonomy
  6. AI hazard list
  7. AI Hazard Management
  8. AI hazard identification
  9. AI risk assessment
  10. AI risk treatment
  11. Documentation
  12. Case Study
  13. Conclusion
  14. Acknowledgement

Figures (2)

  • Figure 1: A taxonomy to classify AI hazards along the axes AI hazard level, AI hazard mode, and AI life cycle stage.
  • Figure 2: A schematic representation of the AIHM framework comprising the three main components: AI hazard identification, AI risk assessment, and AI risk treatment.