What is it about?
Artificial intelligence is rapidly spreading across every industry, making it essential to ensure these systems are safe, fair, and reliable. However, checking AI for hidden biases or technical errors is difficult because these models are so complex. In this paper, we surveyed existing rules, tools, and methods used by experts to inspect AI systems. We discovered that most current approaches miss important steps in how AI is built and used. To fix this, we created a clearer guide that helps organizations pick the exact auditing approach they need based on their size and how they use AI.
Featured Image
Photo by Lukas Blazek on Unsplash
Why is it important?
What sets our survey apart is that it looks beyond generic, one-size-fits-all checklists to address the real-world operational reality of AI. While existing literature focuses heavily on isolated ethical principles, our work maps auditing tools directly across the entire AI value chain and accounts for an organization's specific technical maturity level. Timely context and key differences that drive readership: * Timely Relevance: As global regulations (such as the EU AI Act) transition from high-level proposals into strictly enforced laws, organizations are urgently searching for practical ways to achieve compliance. This paper provides a timely roadmap for navigating those complex legal and technical requirements. * Fills a Critical Gap: We identify where current auditing practices fall short—specifically how standard audits fail to adapt as an AI system matures from a prototype into production. * Practical Impact: By offering context-specific recommendations, our framework translates abstract ethical guidelines into actionable steps. Readers get a pragmatic guide for choosing the right tools for their specific organizational scale and industry context.
Perspectives
Writing this survey was a deeply rewarding experience, as it grew out of our shared frustration with how theoretical most discussions around AI safety and ethics remain. While working alongside my co-authors, we continually asked ourselves how engineers and compliance teams could actually put these principles into practice today. Synthesizing such a fast-moving, multi-disciplinary field was a immense task, but seeing the framework come together to bridge the gap between abstract policy and practical auditing was incredibly fulfilling. My hope is that this work makes AI governance feel far more approachable and actionable for everyone working on the frontlines of technology.
Usman Shahbaz
Macquarie University
Read the Original
This page is a summary of: Auditing Artificial Intelligence Systems: A Survey of Current Frameworks, Principles and Approaches, ACM Computing Surveys, August 2026, ACM (Association for Computing Machinery),
DOI: 10.1145/3844500.
You can read the full text:
Contributors
The following have contributed to this page
