How to apply safety techniques and measures as presented in safety standards when introducing AI

Myklebust, Thor; Stålhane, Tor

Permalink

https://hdl.handle.net/2142/121837 Copy

Description

Title

How to apply safety techniques and measures as presented in safety standards when introducing AI

Author(s)

• Myklebust, Thor
• Stålhane, Tor

Issue Date

2023

Keyword(s)

• Safety
• Artificial intelligence (AI)
• Techniques and measures
• DevOps

Abstract

The focus of this paper is on the investigation of the suitability of several safety standards in ensuring the safe operation of AI systems based on methods, techniques and measures. Safety standards include several methods, techniques, and measures to ensure that risks are avoided, tolerable or are less likely to occur, mitigated or controlled and if hazards occur, their consequences have less impact. Specifically, the paper examines the techniques and measures outlined in the IEC 61508-3, IEC 61508-7, ISO 26262-6, EN 50128, and ISO/IEC 5469 standards. Additionally, the paper suggests ways to customize these methods, techniques and measures to develop safety systems that incorporate AI technology. As AI systems involve training with real-world or test data, it is crucial to consider both the development and operational phases when adapting these techniques and measures. DevOps is a process that combines software development and operations to shorten the systems development life cycle while delivering new features, fixes, and updates frequently and reliably, while also ensuring safety, security, and efficiency. It emphasizes collaboration, automation, and continuous improvement to enable faster and more frequent deployment of software products. This paper proposes adapting these techniques and measures when include AI and tailoring such an approach to DevOps.

Type of Resource

text

Language

eng

Handle URL

https://hdl.handle.net/2142/121837

Owning Collections