Reaction systems are a qualitative formalism for modeling systems of biochemical reactions. They describe the evolution of sets of objects representing biochemical molecules. One of the main characteristics of Reaction systems is the non-permanency of the objects, namely objects disappear if not produced by any enabled reaction. Reaction systems execute in an environment that provides new objects at each step. Causality properties of reaction systems can be studied by using notions of formula based predictor. In this context, we define a notion of opacity that can be used to study information flow properties for reaction systems. Objects will be partitioned into high level (invisible) and low level (visible) ones. Opacity ensures that the presence (or absence) of high level objects cannot be guessed observing the low level objects only. Such a property is shown to be decidable and computable by exploiting the algorithms for minimal formula based predictors.