Very much like spam: filters are good but not 100% good, so there must be a way to look at what the filter has rejected, and allow the reader to judge.
I've had to deal with Microsoft outright dropping everything from a domain (not on this server), so I do know it happens, I'm just not convinced it's necessary.
> I'm just not convinced it's necessary.
Well yes in some special cases dropping mails is not necessary. My mistake to assume your case is one of the common cases.
https://en.wikipedia.org/wiki/Email_spam#Statistics_and_esti...
Google has some very advanced natural language processing technology. Try this Google search: "What year did Neil Armstrong land on Mars?"
https://news.ycombinator.com/item?id=28003635 ("Ask HN: Googler/YTer Able to Help with CIDRAP's Dr Osterholm's YT Strike?")
I fall pretty strongly on the side of combatting misinformation, but I disagree with outright removal of content. I think it should be flagged as misinformation, but left available. Put another way, I'm a fan of labelling, not censoring.
I want transparency. "Our misinformation bot rated this as a 90% chance of being misinformation because XYZ." I bet the only reason we can't have that is because the ML bots suck so much the tech industry is scared to implement any system that might be open to scrutiny or analysis. It's a bit ironic.