The knowledge of the long-term behavior of nuclear waste in anticipation of ultimate disposal in a deep geological formation is of prime importance in a waste management strategy. If phenomenological models have been developed to predict the long-term behavior of these materials, validating these models remains a challenge, when considering the time scale of radioactive decay of radionuclides of environmental concern, typically 10 ⁴ –10 ⁵ yrs. Here we show how natural or archaeological analogues provide critical constraints not only on the phenomenology of glass alteration and the mechanisms involved, but also on the ability of experimental short-term data to predict long-term alteration in complex environments.