Again, I'm only going by the Californian bill, but that one is pretty clear that the person setting up the account should either supply the user's birth date, or the age bracket they are in. There is nothing indicating this should be validated in any way. I'd agree that, if the machine was compromised, and the user's birth date was used, it would be possible to leak that data, but given those preconditions, it would be one of the least interesting things leaked. I'd certainly prefer to just store the age bracket, and have a way for the computer admin update it as the user grows towards their 18th birthday.

That's a thorny question. The main ways we currently got either involve the sites in question collecting personally identifiable information, such as government issued ID, and making a determination as to what to serve based on the information that contains, or the sites adhereing to a voluntary code, such as RTA, to include an identifying header, and parents installing, configuring and maintaining the software or services to restrict access. The former method is obviously dangerous as it requires handing over your ID, and the latter is all voluntary and so there is little impetus to do it, and the complexity will also act as a deterent for parents. Turning it around and just having the computer send a flag for the age bracket gets rid of the need to transmit personally identifiable information and makes the parents' setup job much easier with a one-shot, tick a box and carry on, process.