Micali, Rabin, and Kilian [9] recently introduced zero- knowledge sets and databases, in which a prover sets up a database by publishing a commitment, and then gives proofs about particular values. While an elegant and useful primitive, zero-knowledge databases do not offer any good way to perform updates. We explore the issue of updating zero-knowledge databases. We define and discuss transparent updates, which (1) allow holders of proofs that are still valid to update their proofs, but (2) otherwise maintain secrecy about the update. We give rigorous definitions for transparently updatable zero- knowledge databases, and give a practical construction based on the Chase et al [2] construction, assuming that verifiable random functions exist and that mercurial commitments exist, in the random oracle model. We also investigate the idea of updatable commitments, an attempt to make simple commitments transparently updatable. We define this new primitive and give a simple secure construction.