AI Compliance Playbook: Traditional Risk Controls for Cutting-Edge Algorithms

The corporate use of artificial intelligence and machine learning (AI/ML) skyrocketed during the Covid-19 pandemic, with companies across the economy investing in algorithmic tools to boost capabilities and tackle major problems. Public questions about misuse have accelerated too, with one in the center spotlight: Is your algorithm fair? Companies have responded by adopting AI/ML ethics policies. Yet, leaders from these companies contend that they now need more than an ethics policy and are advocating for government regulation. This three‑part series from ACR’s sister publication the Cybersecurity Law Report, seeks to fill in the current regulatory gap by describing essential steps for an AI/ML compliance program, including adapting 1970s anti‑discrimination practices to meet the future. It also reports on data scientists’ recent compilations of AI/ML failures. Part two will detail seven AI/ML risks that companies now face, and part three will discuss both cutting-edge AI auditing and the venerable “three lines of defense” approach. See ACR’s three‑part series on artificial intelligence for anti‑corruption compliance:  “Foundations” (Oct. 28, 2020); “Building a Model” (Dec. 2, 2020); and “Five Workarounds for Asymmetric Data Sets” (Feb. 3, 2021).

To read the full article

Continue reading your article with an ACR subscription.