If ever there was a job for ML. It seems like text detection and a classifier of what the text is, then blur that part of the photo. Same with brand names, art and logos in TV shows.