Keep less, say the regulators. Keep everything, says the model. Define "it," says no one.
Ironic subject line on file: “Company Resolves Retention Debate By Losing The Person Who Understood It”
Privacy law asks you to keep only what you need, for only as long as you need it. The current appetite for training data asks you to keep everything, forever, in case it is useful later. You cannot reconcile those two instructions without first agreeing what the data is, who owns it, and what it is for — which is exactly the work that keeps getting deferred. The second story explains why it keeps getting deferred: the chief data officer who would own the answer tends to leave before the definition of "customer" does. Accountability that was never written down cannot be inherited.
Privacy law says keep less. The model says keep everything. Nobody wrote down what "it" is.
Minimization is a sentence about purpose. Most firms never finished the sentence.
The chief data officer leaves before the definition of "customer" does.
The role turns over in months. The undefined terms have been here for years.
Minimization and hoarding are both just guesses until someone owns the definition they disagree about.