I don't see how you can avoid human labor for this type of task given the current energy demands of the top AI models.
Once we do reach a point where AI could do the filtering, who is going to draw the limit to where free speech ends? Should they have that much power?