that's a cheap price to pay to get our netflix suggestion, and search result slightly on point.
If confusing people with gorillas is unacceptable then doing any kind of labeling on arbitrary photos is unacceptable, because that's the kind of mistake computer vision algorithms make. Its idea of what people look like is probably something like
Three splotches in a small range of positions relative to each other. Two eyes and a mouth. That's what makes learning so popular, the programs can be really, really dumb, and most of the time they do ok anyway.