I think it depends on some technical specifics, like which meta data was associated with that content, and the degree to which that content was surfaced to user

I think it depends on some technical specifics, like which meta data was associated with that content, and the degree to which that content was surfaced to users that fit the demographic profile of a ten year old child. If your algorithm decides that things in the 90th percentile of shock value will boost engagement to a user profile that can also include users who are ten years old then you maybe have built a negligent algorithm. Maybe that’s not the case in this particular instance but it could be possible.