Comment by exceptione

I am not oblivious to the intent behind this push, but even if you focus solely on the technicalities the idea falls apart. Even with only client-side verification this will be a big privacy intrusion. I see current AI's flag prompts for the most stupid reasons for using words that might possibly occur in non-safe contexts too. The human experience is just too complex for a machine to understand.

To properly assess something, you need to be bodied in reality, being related to the other human in the same human reality. All the datacenters of the world combined will fail the stated objectives, let alone a stupid phone chip. We should not allow computers to take on the role of policing actors in our human reality, because they even can't perform that role faithfully.