Methodologies for AI Assessments, Reviews and Audits

Today we delve into the intricate world of AI assessment, review and audit methodologies, focusing on international frameworks and regulatory approaches. The discussion features experts from the City Bar Presidential Task Force on Artificial Intelligence and Digital Technologies, including Azish Filabi (American College McGuire Center for Ethics and Financial Services), Rim Belaoud (Forensic Risk Alliance), Nikhil Aggarwal (Deloitte Anti Money-Laundering), Lenka Molins (Oxford Internet Institute) and Jerome Walker (Task Force Co-Chair). They explore the definitions, methodologies, and challenges of AI audits across different jurisdictions such as the US, EU, Canada, and the UK, providing perspectives on issues related to methodologies, bias, transparency, and accountability. The episode also covers practical approaches for organizations to review AI models and highlights the importance of robust AI governance in various sectors, including financial services, AML, CFT, fraud, and export controls.

00:00 Introduction to the Podcast 00:50 Overview of AI Assessments, Reviews, and Audits 02:20 Key Definitions and Concepts in AI 05:44 Panelist Introductions 08:39 Discussion on Responsible and Trustworthy AI 18:33 Training AI Models and Explainability 22:33 Challenges in AI Assessments and Reviews 27:09 Global Perspectives on AI Audits 39:10 Practical Approaches for AI Model Reviews 53:57 Key Skills for AI Model Audits 59:27 Introduction and Areas of Practice 01:01:31 AI in Anti-Money Laundering and Counter-Terrorist Financing 01:07:36 AI Models in Fraud Detection 01:14:41 Export Control on AI Models 01:21:35 International AI Audit Methodologies 01:27:42 Challenges in AI Audits 01:42:10 Accountability in AI Audits 01:46:13 Conclusion and Final Thoughts