Much of the current problem is that fake news spreads faster than moderators can make a decision on it, or journalists can fact-check it. If you can keep it in a "slow burn" phase longer, where it's being forwarded along to a handful of people at a time, it's easier to combat.
The point is, this has next to zero selectivity. It just makes the platform hostile to encrypted group messages. My guess would be that (besides PR regarding the Indian stories, which is a social problem, not a technical one) this is about censorship. You can't censor things you can't see. Better stop them from communicating freely.
This seems like it could only be implemented client-side. It also seems like it would require rapid and frequent updates to the ever-growing blacklist. That in turn would require a way to figure out what hashes needed to be blacklisted, which implies either less-than-ideal cryptography practices or the ability to decrypt messages. Neither is desirable for user privacy.
I have a rule I use in computing: whenever you think you want a blacklist, you probably want a whitelist.