I've used Google's Cloud Vision API to detect explicit content, and found it quite effective at flagging all kinds of risky stuff. (And a unique challenge for testers on how to exactly to safely test for such content.)