It is rather interesting to note that this Corpus of data may not be as valuable if it cannot be used without always being legally in several grey areas (perhaps even red areas in some jurisdictions).

Currently, an increasingly large pool of artist/writters/singers and other people (even corporations such as studios and large right holders) are exercising their rights to not have their creations and derived works be used or slurped into AI models without their express consent.

Corporations making use of those AI models may find themselves in expensive legal limbo now and the foreseeable future.

Considering no redditor imagined nor consented to have their post and comment history be comprehensively abused (as in "improper treatment or usage; application to a wrong or bad purpose; an unjust, corrupt or wrongful practice or custom").

We may enter a period where lawlessness pervades AI models (just like any gold rush, for example the current crypto craze). Eventually, the legal framework will catch up and will probably make any dubious Corpus of data untouchable.

How long this takes is anyone's guess. I surmise several large profile lawsuits would suffice.