Seems like a poorly trained ML model - crudely speaking, perhaps its training set was tilted towards descriptions of conservative news sites so it learned to insert the word “conservative” when describing a news site.