I think that the improved version of age verification is to ask the yes/no question to a government third party based on a signed payload that your local device offers the service. The government already has your identifying data, they only need to certify on behalf of which person the question is asked.
Which a) has a whole host of other concerns associated with it, and b) still does not solve that problem, because it's not at all hard for a child (especially a teenager!) to sneak their parent's ID, use it to authenticate for a service, then put it back.
After all, are most services going to require the ID to be present for every session? Or are they going to require a one-time authentication for the account?
