The use of animals to model humans in biomedical research relies on the notion that basic processes are sufficiently similar across species to allow extrapolation. Animal model validity is discussed in terms of the similarity between the model and the human condition it is intended to model, but no formal validation of models is applied. There is a stark contrast here with the use of non-animal alternatives in toxicology and safety studies, for which an extensive validation is required. We discuss both the potential and the limitations of validating preclinical animal models for proof-of-concept studies, by using an approach similar to that applied to alternative non-animal methods in toxicology and safety testing. A major challenge in devising a validation system for animal models is the lack of a clear gold standard with which to compare results. While a complete adoption of the validation approach for alternative methods is probably inappropriate for research animal models, key features, such as making data available for external validation and defining a strategy to run experiments in a way that permits meaningful retrospective analysis, remain highly relevant.