From the article...
But while many think that YouTube's system isn't great, Trendacosta also said that she "can't think of a way to build the match technology" to improve it, because "machines cannot tell context." Perhaps if YouTube's matching technology triggered a human review each time, "that might be tenable," but "they would have to hire so many more people to do it."
That's what it comes down to, right there.
Google needs to spend money on people, and not just rely on the AI automation, because it's obviously getting things wrong, its not judging context correctly.