It's the Scunthorpe problem all over again. It's like Youtube flagging that chess channel for hate speech[0] (perhaps because words like "black", "white" and "attack" were used a lot). It will annoy people who're accidentally caught by the filter, and people who want to get past it will figure out a way anyway (by look-alike character substitution or whatever). I think it's the wrong approach to that problem.
Indeed it is. There are comments here and there in the GitHub issues where the devs defend the filter as being targeted at making it harder for "the [American] right wing" to use the platform.
[0]: https://news.ycombinator.com/item?id=27914261