From Wild West to Collective Vision: Navigating the Perils of Audit Washing in the AI Auditing Ecosystem

AI audits comprehensively examine applications against a set of rules, data quality, model bias, and emergency procedures, among others. If successful, the auditor certifies the application’s trust. Contrary to the market for financial audits, the ecosystem for AI audits is quasi-non-existent: In this unregulated ‘Wild West’ there are almost no rules, very few auditors, and unaware auditees. However, we should care about ensuring meaningful AI audits that take into account the voices of civil society and companies. 

Following a brief 101 on AI audits, participants take the roles of policymakers, auditors, or auditees, and formulate their demands. How do we ensure professional standards that prevent a race to the bottom? How can audits be accessible to SMEs? How can auditees find their best auditing partners? Our session aims to form a shared vision of a trustworthy AI auditing ecosystem.
 
Such a vision is urgently needed, as German policymakers prepare to implement the EU AI Act, which foresees a market for AI audits but remains vague on details.