Auditing the Black Box: Rethinking Audit Assurance and Professional Liability in the Age of Artificial Intelligence

Abstract

Artificial intelligence has moved from the periphery of audit practice to its operational core, and the profession’s conceptual apparatus has not kept pace. This article examines what the opacity of machine learning systems does to two foundations of the auditing discipline: the assurance model, which rests on the auditor’s ability to obtain and evaluate sufficient appropriate evidence, and the liability regime, which rests on a standard of care calibrated to human judgment. Drawing on auditing scholarship, standard-setting developments at the IAASB and the PCAOB, comparative regulatory instruments including the European Union’s Artificial Intelligence Act, and the common law of auditor negligence, the article argues that neither wholesale prohibition of opaque systems nor uncritical reliance on them is defensible. It proposes a graduated algorithmic reliance framework that ties the permissible depth of reliance on an AI system to the demonstrable explainability of that system, the materiality of the assertion it supports, and the auditor’s capacity to corroborate its output through independent means. The article also reformulates the negligence standard around the figure of the competent hybrid auditor and considers how liability should be allocated among audit firms, technology vendors, and audited entities. Particular attention is given to the position of developing economies, where regulatory capacity constraints sharpen every one of these questions.