Computer predictions are now being accepted for some purposes in submissions to regulatory bodies for the commercial use of pharmaceuticals and other chemicals. It is notoriously difficult to repeat computer modelling experiments on different sites and, especially, at times long after the original one. So it may be only a matter of time before a scandal emerges when a chemical is found to be unsafe and the study that led to its having been claimed to be safe cannot be repeated. Outline guidelines for good computer modelling practice are proposed.