The development of networks connecting a large number of distributed databases will shortly make available extremely large amounts of data. For instance there is currently a project whose objective is the interconnection of one hundred existing molecular-biology databases. This kind of interconnection raises two issues: How is relevant information to be retrieved ? And what is the validity of information obtained from different sources, assuming that some of them are reliable and others are not ? These issues are much more important in this context than in the context of accessing a unique database because each database has been designed independently, and they are also updated independently. This abstract presents some attempts to provide solutions to these two problems in the uniform formalism of Modal Logic. The main results, at the present time, are at the theoretical level. Nevertheless at the end of the abstract some guidelines are described which indicate how these results could be used in potential implementations.