The aim is to redefine AI auditing by emphasizing the conflicting interests between auditors and AI models.
Conceptual analysis of auditing frameworks in AI.
Exploration of adversarial relationships between models and auditors.
Evaluation of existing practices in AI auditing.
Identified limitations in traditional auditing methods due to black-box assumptions.
Proposed a framework that considers auditor motivations and AI model defenses.
Highlighted the need for transparent processes in AI assessments.
Abstract
Shifting the central focus in AI auditing from assuming truthful model responses to considering the antagonistic interests of an auditor and the AI model being scrutinized.