It's extremely difficult, and as I said it will be easily abused by any government which is looking to manufacture consent among its population.
One person's definition of harm is completely different than another person's.
>The argument that words can harm has also been used to justify institutional efforts to prohibit certain language. In 2022, the Stanford University IT department launched a program entitled the “Elimination of Harmful Language Initiative,” which proposed removing from Stanford's websites a set of words considered to be biased or discriminatory, including “blind review,” “handicap parking,” “OCD,” “victim,” “immigrant,” and “American”
>https://www.sciencedirect.com/science/article/pii/S019188692...
What happens when we get an administration that decides a platform which allows "discriminatory" language should be punished via this new "oversight" mechanism?