ITI Policy Principles for Enabling Transparency of AI Systems

The purpose of this document is twofold – first, to educate policymakers about transparency in the context of AI systems, and second, to offer suggested policy approaches to facilitating greater transparency of AI systems. We stress that the better policymaking approach is to apply transparency requirements to specific, highrisk uses of AI systems, as opposed to imposing requirements on the transparency of algorithms which, although an important part of AI systems, are also an integral component of a much broader set of technologies used in a wide variety of settings. Indeed, it is the context in and purpose for which the algorithm is used that matters—not the fact that an algorithm is used in and of itself—and policymakers should recognize that there are many contexts in which providing transparency is likely not necessary.