On the plus side, this is the kind of stuff you could screen pretty easily with large model machine learning. Not that there is a business in identifying scientific fraud, doing that with fraudulent government documents would probably have a better ROI (at least for the tax payer), but clearly we need a repository if every image/graph that has been published as evidence to start.

It would be something you could offer to journals perhaps as a business. Sort of "peer reviewed and fraud analyzed" kinda service.

What is truly sad for me is the 'wrong paths' many hard working and well meaning scientists get deflected down while someone cheats to get more 'impact' points.