Artificial intelligence or “AI” encompasses a wide variety of technologies. We focus on “predictive algorithms”, an important class of algorithms that includes machine learning algorithms. The general concept of a “predictive algorithm” is useful for many regulation/oversight purposes, and covers a useful subset of the algorithms referred to as “AI” in recent public discourse. The use of predictive algorithms within the New Zealand government sector is not a new phenomenon. Algorithms such as RoC*RoI in the criminal justice system have been in use for two decades. However, the increasing use of these tools, and their increasing power and complexity, presents a range of concerns and opportunities. The primary concerns around the use of predictive algorithms in the public sector relate to accuracy, human control, transparency, bias and privacy.