I'm following up after some testing I've done; it looks as though only the 1st frame of an animated GIF is assessed by the content safety API, which isn't ideal!
I'd like someone from the Microsoft team to take this on as an improvement request; it really should go beyond just the 1st frame, possible unpacking it into a sprite map if internally it doesn't want to analyse each frame individually.
Can anyone from MS comment on this please?