[D] Can we define the expiry date of a Machine Learning model?

There is a lot of literature on model drift, data drift, model retraining etc. But can we really define an expiry date for an ML model? If so what are the factors it will depend on?

learning machine machine learning machinelearning