Lol. Whoever tuned that model has never heard of recall and their training set probably was lots of p*rn and p*rn only.
That is exactly why you should do a decent ML course about model evaluation practices and ethical implications before you start training models.
I tested a machine learning model designed to detect NSFW content. It marked this image as unsafe, because of Tim Allen's breasts (marked in blue) and my hand-penis (marked in green)