All the numbers he's shown are probably bollocks - how can you actually know how much percentage of "hate speech" you capture? You can say we have "blocker or removed x% more messages than y months ago" - but "the amount of hate speech we detect proactively, before anyone reports it, has more than doubled from 24% to 52%" - how did you identify the "100%" baseline? If you know that's 52% of something, you evidently know enough about what the remaining 48% is... and why you didn't act on it?
Or they meant they identified it but let it through until someone reported it, so they caught only 52% of reports proactively? But why let it through if you have identified it??? How many times it was used before being reported????
It looks to me just numbers made up by some PR team and thrown there just because they look good.
While, how that could be confirmed independently? Did they got only the low-hanging fruits to increase some number, maybe letting thorough the most dangerous ones? We are not speaking about identical items - there's a big difference among these kind of topics.