It's not a guarantee, but generally, digital ID systems live on the phone of one person only and require a screen lock to use. You're right though, that there is nothing preventing someone else from borrowing another's identity.

It was really hard for me not to claim I'm a minor on Lemmy NSFW, I knew I should not do it, after getting naked on there, would cause TOO MUCH chaos.

It's because it's not about age verification, it's about surveillance.

With something like a physical gift card.

Go to a store or kiosk, show them your ID card or driver's license, and they'll give you a card randomly chosen from the shelf with a code to activate the +18 version of any social network of your choice.

Each code could only be used once. People would have to buy more, at a symbolic cost, for each social network they wished to activate.

I would tend to be against this in the same way, but at least it would be something I could understand where the objective is actually what is being presented (protecting the children), albeit misguided, because to me it is clear that what is currently being promoted and proposed has nothing to do with age verification, but rather with mass surveillance, marketing and censorship.

In general, we accept that the Government already knows who you are, how old you are, and where you live. That's already a given. The purpose of a zero-knowledge age verification scheme is to allow a third party (not the Government) to be confident that a person is an adult, without being given any additional information or being able to deduce any additional information from what they're given. So essentially, they get only 1 bit of information: whether the user is an adult (true/false). In practice, a perfect system is not possible, since the fact that you receive a response also means you get the answer to related questions, like whether the user possesses a Government-issued ID (obviously "true" if they can successfully complete the verification).

So, here's how such a scheme might work. There are many possible implementations.

In the United States, we have (optional) digital ID cards. These are added to one's digital wallet in a similar manner to payment cards and can be used for things like buying alcohol, getting through airport security, and driving. This digital infrastructure can be re-used.

1. An organisation which wants to perform digital identity verification generates a cryptographic key pair and registers the public key with a Government server ahead of time. The public key is published to a Government-run public keyserver.
2. A website who wants to verify a user's age sends a verification request to a Government server, digitally signed with their private key. The server responds with a request ID, which is a random, but unique, string of characters.
3. The website provides this string to the user. The user copies the string.
4. The user opens their digital wallet, selects their ID card, and then opens the age verification feature. The user pastes the request ID into their digital wallet, which fetches information about the request from the Government server. Because the request which the request ID is associated with was signed using the organisation's private key, the Government can tell the user who initiated the request.
5. The user is asked to confirm/deny the age verification request. If the user confirms the request, then a biometric will be required to access their private key (these are stored in the device's keystore), sign the approval response, and then sent that response to the Government server. The Government server checks that the signature is valid and tied to the key associated with that ID before marking the verification request as completed.
6. After confirming, the user returns to the website and clicks a button which says "I've completed the verification." The website then queries the request ID with the Government server (again, signing the request with their private key). The Government server responds with "completed" if the user has accepted the request, or "not completed" if the user has either not yet accepted the request or denied it.

It is possible to construct a zero-knowledge proof using cryptography and adapting existing digital ID infrastructure. A user can prove that they have knowledge of a private key tied to an adult's identification card without having to reveal the key, or the associated public key.

But that being said, whether something is possible and whether it is a good idea are two different questions.

Actually its possible, but we do not want that

What do you mean? The system already exists in Germany and whatever website or app asking if you‘re an adult will only get exactly that information. Not your face, not your name, not even your age. They will simply receive a verified Y/N from a government service. Needless to say no US company implemented it because they would miss out on a lot of sweet sweet data.

It already exists, lots of institutions people work with know way more than just their age. They could really provide an anonymous way of validating someone's age without divulging their identity. This can also be done in such a way that the verification provider doesn't know the requester either so they don't have a way of tracking user habits

"Open weights" isn't worth a shit either.

If you don't have the access to training data or the computational power to bake it into a model, you are beholden to somebody's (or rather, some corporation's) binary blob. We're talking about the difference between freeware and FOSS effectively.

Not that it matters because all of thin generative AI stuff is for talentless tech bro chuds in the first place...

When asked if he prefers the human race continues existing or not Peter Thiel hesitated and was like “er uhh well”. That should tell any regular person what type of ghoul he is.

Yeah boii, I wanted to use someone's concoction of Qwen3.5 and Claude, but it wasn't compatible with Ollama for some reason. Link

Oh yeah, I too am not happy about the tech world calling this creation AI when it doesn't mimic thinking for itself -- it's still just working on strict parameters under variable parameters, it's just code, and as such performs pretty much every task poorly compared to humans*. 'AI' was supposed to be reserved for what I suppose is now AGI.

*Unless you pour massive money, resources, energy and labour into making a model that's still as bad, but much faster and pretends to be an AGI