The whole point is that they're creating a proprietary dataset. If they re-sold it, then other people could use that data.
Hopefully the scans will stick around and when these companies are liquidated they aren't lost.
The only contribution that LLMslop companies will make to human history is indexing all available human knowledge down to the letter. Opening those datasets to the public would be a social good that would almost offset the damage they've done (assuming they collapse soon).