Seems like they've tried to roll their own machine learning classifier as well because I've seen some comically bad screenshots of banned content. For example, a picture of someone's finger got banned as sexually explicit.

Should have just used AWS or Google Cloud AI APIs