Operationalising AI governance through ethics-based auditing: An industry case study

Ethics-based auditing (EBA) faces practical governance challenges beyond technical metrics: The study found that the main difficulties organizations encounter when conducting AI ethics audits are not primarily about developing better evaluation methods, but rather classical governance issues such as maintaining consistent standards across different departments, clearly defining audit scope, and managing internal communication—suggesting that implementation challenges are organizational rather than purely technical.

Large multinational organizations struggle with decentralization during AI audits: AstraZeneca's 12-month audit revealed that ensuring harmonized ethical standards across a decentralized organization is a significant barrier to effective EBA, indicating that companies with multiple divisions or global operations face particular challenges in operationalizing AI ethics principles uniformly.

Measuring actual outcomes of ethics audits remains difficult: The research highlights that organizations cannot easily demonstrate whether ethics-based audits actually lead to improved AI practices, suggesting that current approaches may lack clear mechanisms for tracking whether audit recommendations translate into meaningful behavioral change.

Organizational context is critical for successful AI ethics audits: The case study emphasizes that EBA procedures must be carefully integrated into an organization's existing structures and processes to be both feasible and effective, rather than being implemented as standalone technical exercises.

Gap remains between AI ethics principles and real-world practice: While EBA is promoted as a tool to bridge the gap between stated ethical principles and actual implementation, this study suggests the challenge is more complex than previously acknowledged in academic literature, requiring deeper attention to organizational change management.