Danny Werfel (Duke) presents Risk Framework For The Use Of Artificial Intelligence In Tax Administration and Practice at Duke today, as part of its Tax Policy Seminar hosted by Larry Zelenak:

This tax policy seminar will focus on a potential new risk framework for the responsible use of artificial intelligence (AI). In conversations with tax practitioners, administrators, and advisors, a consistent theme has emerged: while interest in AI adoption is accelerating, there is not yet a widely accepted, trusted framework to help organizations identify, evaluate, and mitigate the associated risks.

This author intends the risk framework to help prevent rapid integration of AI outpacing necessary guardrails. The proposed framework introduces two distinct risk registers: one designed for tax preparers (“Stack Model”) and another for tax administrators (“Quadrant Model”). While these registers reflect the unique operational challenges faced by the public sector vs. the private sector, they also reveal important areas of overlap, particularly in data integrity, algorithmic bias, and the protection of taxpayer rights.