Perfect enforcement of rules is rarely a goal of policy. It just sets expectations for behavior, and when violation of policy becomes known remedial and/or punitive measures can be taken.
> Should we roll back those laws? Fraud is fine as long as the party defrauded can't find out?
I think you're putting words into my mouth.
Anyway, what's extra difficult is that when using data to build an ML model, you can actually throw away the data afterwards, like source code after compiling a program.
Perfect enforcement of rules is rarely a goal of policy. It just sets expectations for behavior, and when violation of policy becomes known remedial and/or punitive measures can be taken.