I think we're looking for racism where there is none. Observing physical characteristics is not racism, and the fact of the matter is that some black people tend to have a more protruding jawline and fuller lips than one would see in a specific group of people who are caucasian. If you compare those two, very specific, physical characteristics with the great ape family, you see similarities.
Now, before anyone starts screaming about how I'm racist (too late), having one or two physical characteristics with another species, out of hundreds of physical characteristics, doesn't mean a damn thing when it comes to humans easily identifying people as not belonging to the same subgroup that shares those characteristics. It's the same reason why we don't think that someone with a striped mohawk is a zebra. We're able to take contextual clues and infer that that person is not, in fact, a zebra.
However, it DOES mean that an algorithm that has been largely trained on either insufficient or faulty data sets can make incorrect inferences based on the characteristics that it has been trained to recognize. If anybody is at fault here, it isn't the algorithm, but whatever engineer fed their image recognition an algorithm so woefully insufficient that it would confuse a human with an ape just based off of a physical characteristic or two out of any number of data points that would indicate, "Hey, this is obviously a human being".
Having done some random searches I agree this is just a mistake - and comparable to others. A few examples that I have found are
- a search for "dolphins" including a picture of my daughter swimming.
- a search for "squirrel" including meekcats
- a search for "cat" including some dogs
- a search for "ghost" including a slightly blurred picture of my wife
- a search for "man" showing some women and "woman" showing some men