Attempting to handle prompt injections by prompting the model (not to leak sensitive data), is like attempting to stop a fire by burning the area around it
Age Verification is not about “saving kids from spying/social media”.
Same as Regulatory Encryption Backdoors are not about “protecting kids from child predators”
This is framing.
Abusing kids is bad. But that is not what this about. Let’s keep the discussion on topic: the goal is 1984-level oversight… not just of the kids
[Edit: I guess HN agrees: meanwhile the title changed from “saving kids from spying” to “is actually mass-surveillance” - good]
So all we need is ‘controlled prompting’ to handle prompt injections :-)